Embed Size (px)
Citation preview
DECISION PERFORMANCE USING SPATIAL DECISION SUPPORT SYSTEMS:
A GEOSPATIAL REASONING ABILITY PERSPECTIVE
by
MICHAEL A. ERSKINE
B.S., Metropolitan State University of Denver, 2004
M.S., University of Colorado Denver, 2007
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Computer Science and Information Systems
2013
ii
This thesis for the Doctor of Philosophy degree by
Michael A. Erskine
has been approved for the
Computer Science and Information Systems Program
by
Jahangir Karimi, Chair
Dawn G. Gregg, Advisor
Judy E. Scott
Ilkyeun Ra
November 8, 2013
iii
Erskine, Michael, A., (Ph.D., Computer Science and Information Systems)
Decision Performance Using Spatial Decision Support Systems:
A Geospatial Reasoning Ability Perspective
Thesis directed by Associate Professor Dawn G. Gregg.
ABSTRACT
As many consumer and business decision makers are utilizing Spatial Decision
Support Systems (SDSS), a thorough understanding of how such decisions are made is
crucial for the information systems domain. This dissertation presents six chapters
encompassing a comprehensive analysis of the impact of geospatial reasoning ability on
decision-performance using SDSS. An introduction to the research is presented in
Chapter I. Chapter II provides a literature review and research framework regarding
decision-making using geospatial data. Chapter III presents the constructs of geospatial
reasoning ability and geospatial schematization. Chapter IV explores the impact of
geospatial reasoning ability on the technology acceptance of SDSS. Chapter V presents
results of an experiment exploring the impact of geospatial reasoning ability on decision-
making performance. Finally, Chapter VI presents a conclusion. Together these chapters
contribute to a greater understanding of the impact of geospatial reasoning ability in
relation to business and consumer decision-making.
The form and content of this abstract are approved. I recommend its publication.
Approved: Dawn G. Gregg
iv
TABLE OF CONTENTS
CHAPTER
I. AN INTRODUCTION TO DECISION-MAKING USING GEOSPATIAL DATA ..... 1
Abstract ................................................................................................................... 1
Introduction ............................................................................................................. 1
II. BUSINESS DECISION-MAKING USING GEOSPATIAL DATA: A RESEARCH
FRAMEWORK AND LITERATURE REVIEW ............................................................... 6
Abstract ................................................................................................................... 6
Introduction ............................................................................................................. 7
Literature Review.................................................................................................. 11
Theoretical Background .................................................................................. 11
Information Presentation ................................................................................. 12
Task Characteristics ........................................................................................ 16
User Characteristics ........................................................................................ 23
Decision-Making Performance ....................................................................... 30
Conceptual Model ................................................................................................. 33
Discussion ............................................................................................................. 35
Limitations of Reviewed Literature ................................................................ 35
Future Research .............................................................................................. 37
Conclusion ............................................................................................................ 39
III. GEOSPATIAL REASONING ABILITY: CONSTRUCT AND SUBSTRATA
DEFINITION, MEASUREMENT AND VALIDATON ................................................ 42
Abstract ................................................................................................................. 42
Introduction ........................................................................................................... 43
Literature Review.................................................................................................. 45
v
Cognitive Fit Theory ....................................................................................... 45
Geovisualization and Decision-Performance .................................................. 46
User-Characteristics ........................................................................................ 47
Task-Characteristics ........................................................................................ 49
Scale Development Procedure .............................................................................. 51
Construct Conceptualization (Step 1) ................................................................... 53
Factor One: Examination of Prior Research ................................................... 53
Factor Two: Identification of Construct Properties and Entities .................... 55
Factor Three: Specification of the Conceptual Theme ................................... 56
Factor Four: Definition of the Construct and Substrata .................................. 56
Measurement Item Generation (Step 2) ................................................................ 58
Content Validity (Step 3) ...................................................................................... 63
Measurement Model Specification (Step 4) ......................................................... 67
Reflective and Formative Constructs .................................................................... 69
Visual Representation of Measurement Model ..................................................... 70
Mathematical Notation of Measurement Model ................................................... 71
Data Collection (Step 5)........................................................................................ 72
Pretest A .......................................................................................................... 72
Pretest B .......................................................................................................... 73
Scale Purification and Refinement (Step 6) .......................................................... 73
Pretest A .......................................................................................................... 73
Pretest B .......................................................................................................... 74
Pretest B – Reduced Items .............................................................................. 82
Data Collection and Reexamination of Scale Properties (Step 7) ........................ 87
Collinearity ..................................................................................................... 94
vi
Gender Concerns ............................................................................................. 94
Discussion ............................................................................................................. 96
Implication to Research .................................................................................. 99
Implication to Industry .................................................................................... 99
Limitations and Future Research .................................................................. 100
Conclusion .......................................................................................................... 104
IV. USER-ACCEPTANCE OF SPATIAL DECISION SUPPORT SYSTEMS:
APPLYING UTILITARIAN, HEDONIC AND COGNTIVE MEASURES ................. 106
Abstract ............................................................................................................... 106
Introduction ......................................................................................................... 106
Perceived Enjoyment: Hedonic Measures .................................................... 108
Technology Acceptance Model: User-Acceptance and Utilitarian Measures
....................................................................................................................... 109
Geospatial Reasoning Ability: Cognitive Measures ..................................... 110
Research Model .................................................................................................. 111
Research Methodology ....................................................................................... 117
Research Sample ........................................................................................... 117
Research Instrument...................................................................................... 118
Analysis............................................................................................................... 122
Discussion ........................................................................................................... 131
Implications for Industry............................................................................... 133
Implications for Scholarship ......................................................................... 134
Limitations .................................................................................................... 134
Future Research ............................................................................................ 135
Conclusion .......................................................................................................... 135
vii
V. INDIVIDUAL DECISION-PERFORMANCE OF SPATIAL DECISION SUPPORT
SYSTEMS: A GEOSPATIAL REASONING ABILITY AND PERCEIVED TASK-
TECHNOLOGY FIT PERSPECTIVE ........................................................................... 137
Abstract ............................................................................................................... 137
Introduction ......................................................................................................... 138
Literature Review................................................................................................ 139
Cognitive Fit Theory ..................................................................................... 140
Decision Performance ................................................................................... 141
Perceived Task-Technology Fit .................................................................... 141
User Characteristics ...................................................................................... 142
Task Characteristics ...................................................................................... 143
Research Model .................................................................................................. 146
Research Methodology ....................................................................................... 149
Experiment Design........................................................................................ 151
Subjects ......................................................................................................... 151
Geospatial Reasoning Ability Measurement Items....................................... 152
Perceived Task-Technology Fit Measurement Items ................................... 153
Problem Complexity ..................................................................................... 154
Visualization Complexity ............................................................................. 155
Analysis............................................................................................................... 156
Measurement Model ..................................................................................... 156
Structural Model ........................................................................................... 161
Heterogeneity ................................................................................................ 164
Findings and Discussion ..................................................................................... 166
Limitations .................................................................................................... 168
Implications................................................................................................... 170
viii
Conclusion .......................................................................................................... 171
VI. CONCLUSION: TOWARD A COMPREHENSIVE UNDERSTANDING OF
GEOSPATIAL REASONING ABILITY AND THE GEOSPATIAL DECISION-
MAKING FRAMEWORK ............................................................................................. 172
Abstract ............................................................................................................... 172
Introduction ......................................................................................................... 172
Conceptual Model ............................................................................................... 173
Implication to Research ................................................................................ 175
Implication to Industry .................................................................................. 175
Theoretical Frameworks ..................................................................................... 176
Findings and Discussion ..................................................................................... 178
Future Research .................................................................................................. 182
REFERENCES ............................................................................................................... 184
ix
LIST OF TABLES
Table
1. Common Theories Related to Spatial Decision-Making. ............................................. 12
2. List of Complexity Frameworks (Adapted from Gill and Hicks, 2006). ...................... 20
3. User-Characteristic Measurement Instruments used in Examined Research, not
including spatial ability measures (as shown in Table 4). ................................................ 27
4. Spatial Reasoning Instruments Used in Examined Research. ...................................... 30
5. Research Population Groups. ........................................................................................ 36
6. Spatial Reasoning Instruments Used in Examined Research ....................................... 49
7. Proposed Substrata of GRA. ........................................................................................ 57
8. Definition of SPGS. ..................................................................................................... 58
9. Initial Self-Perceived Geospatial Orientation and Navigation Measurement Items. ... 60
10. Initial Self-Perceived Geospatial Memorization and Recall Measurement Items. .... 61
11. Initial Self-Perceived Geospatial Visualization Measurement Items. ........................ 62
12. Initial Self-Perceived Geospatial Schematization Measurement Items. .................... 63
13. Content Validity Results of Geospatial Orientation and Navigation. ........................ 64
14. Content Validity Results of Geospatial Memorization and Recall. ........................... 65
15. Content Validity Results of Geospatial Visualization. .............................................. 66
16. Content Validity Results of Geospatial Schematization. ........................................... 66
17. Pretest A Reliability Statistics. ................................................................................... 74
18. Pretest B Descriptive Statistics of Demographics Variables. ..................................... 76
19. PLS Factor Analysis Loadings................................................................................... 79
20. Latent Variable Correlation. ...................................................................................... 80
21. Path Coefficients. ....................................................................................................... 80
x
22. Construct Reliability. ................................................................................................. 81
23. PLS Factor Analysis Loadings................................................................................... 83
24. PLS Latent Variable Correlation. .............................................................................. 83
25. Path Coefficients. ....................................................................................................... 84
26. Construct Reliability. ................................................................................................. 85
27. PLS Factor Loading and Cross Loading. ................................................................... 86
28. Inter-Construct Correlations and Square Root of AVE. ............................................ 87
29. Study 1 Descriptive Statistics of Demographic Variables. ........................................ 88
30. PLS Factor Analysis Loadings and Weights. ............................................................ 90
31. Latent Variable Correlation. ...................................................................................... 90
32. Path Coefficients. ....................................................................................................... 91
33. Construct Reliability. ................................................................................................. 92
34. PLS Factor Loading and Cross Loading. ................................................................... 93
35. Inter-Construct Correlations and Square Root of AVE. ............................................ 94
36. SPGON GRA: Gender Group Differences. ........................................................... 95
37. SPGMR GRA: Gender Group Differences. .......................................................... 95
38. SPGV GRA: Gender Group Differences............................................................... 96
39. Final GRA Measurement Items. ................................................................................ 98
40. Final SPGS Measurement Items. ............................................................................... 98
41. Perceived Enjoyment Measurement Items. .............................................................. 118
42. Perceived Usefulness Measurement Items. .............................................................. 119
43. Perceived Ease-of-Use Measurement Items. ........................................................... 120
44. Attitude Measurement Items. ................................................................................... 120
45. Behavioral Intent Measurement Items. .................................................................... 121
46. Geospatial Reasoning Ability Measurement Items.................................................. 122
xi
47. Descriptive Statistics of Demographic Variables. ................................................... 123
48. Item Reliability. ....................................................................................................... 124
49. Item Loadings and Cross Loadings, highest loadings shown in bold. ..................... 126
50. Square Root of Loadings and Cross-Loadings, highest loadings shown in bold..... 127
51. Inter-Construct Correlations and Square Root of AVE. .......................................... 128
52. Construct AVE. ........................................................................................................ 129
53. Path Coefficients. ..................................................................................................... 131
54. Hypothesis Tests. ..................................................................................................... 131
55. Summary of Geospatial Decision Performance Research. ...................................... 145
56. Summary of Geospatial Decision Performance Research. ...................................... 150
57. Descriptive Statistics of Experiment Subjects. ........................................................ 152
58. Geospatial Reasoning Ability Measurement Items, Adapted from Erskine and Gregg
(2011, 2012, 2013). ......................................................................................................... 153
59. Perceived Task-Technology Fit Measurement Items, Adapted from Karimi et al.
(2004) and Jarupathirun and Zahedi (2007).................................................................... 154
60. Problem Complexity ................................................................................................ 155
61. Cronbach’s alpha and Composite Reliability. ......................................................... 157
62. Measurement Item Loadings.................................................................................... 158
63. Average Variance Extracted (AVE) by Construct. .................................................. 159
64. Measurement Item Cross-Loadings. ........................................................................ 160
65. Latent Variable Correlations and Sq. of AVE (shown in bold). .............................. 161
66. R2 Values of Endogenous Latent Variables. ............................................................ 161
67. Path Coefficients and Significance Levels. ............................................................. 162
68. Path Coefficients and Significance Levels. ............................................................. 163
69. R2 and Q
2 Values of Endogenous Latent Variables. ................................................ 164
70. Path Coefficients – Male Only. ................................................................................ 165
xii
71. Path Coefficients – Female Only. ............................................................................ 165
72. Result of Gender Group Comparison. ..................................................................... 166
73. Hypotheses Test. ...................................................................................................... 168
xiii
LIST OF FIGURES
Figure
1. Proposed Research Model............................................................................................. 34
2. Overview of Scale Development Procedure from MacKenzie et al. (2011). .............. 52
3. First-Order Reflective, Second-Order Formative Measurement Model including
Propositions....................................................................................................................... 71
4. Pretest B, Path Analysis including SPGS. ................................................................... 77
5. Pretest B, Path Analysis including SPGS. ................................................................... 78
6. Pretest B, Path Analysis. .............................................................................................. 82
7. Test, Path Analysis....................................................................................................... 89
8. Proposed Research Model.......................................................................................... 112
9. Initial Nomological Network of GRA, along with Path Coefficients and R Square
Values. ............................................................................................................................ 129
10. Proposed Research Model........................................................................................ 146
11. Experiment Workflow. ............................................................................................ 150
12. “Apartment Finder” Experiment Tool Developed using GISCloud and Bing Maps.
......................................................................................................................................... 151
13. Visual result of SEM-PLS Algorithm (using SmartPLS). ....................................... 162
14. Visual result of SEM-PLS Algorithm (using SmartPLS). ....................................... 163
15. Research Model with Relationship Significance. .................................................... 167
16. Conceptual Model. ................................................................................................... 174
17. Current Nomological Network of GRA Construct. ................................................. 182
xiv
LIST OF ABBREVIATIONS
ATM Automated Teller Machine
AVE Average Variance Extracted
CFT Cognitive Fit Theory
DSS Decision Support System
GIS Geographic Information System
GPS Global Positioning System
GRA Geospatial Reasoning Ability
INSPIRE Infrastructure for Spatial Information in the European Community
IS Information Systems
NASA National Aeronautic and Space Administration
NFC Need for Cognition
NSDI National Spatial Data Infrastructure
PDA Personal Digital Assistant
PLS Partial Least Squares
PPGIS Public Participation Geographic Information Systems
PTTF Perceived Task-Technology Fit
RFID Radio Frequency Identification
SDSS Spatial Decision Support System
SEM Structural Equation Modeling
SMW Subjective Mental Workload
SPGON Self-Perceived Geospatial Orientation and Navigation
SPGMR Self-Perceived Memorization and Recall
SPGV Self-Perceived Geospatial Visualization
SPGS Self-Perceived Geospatial Schematization
TAM Technology Acceptance Model
TLX Task Load Index
USGS United States Geological Survey
UTAUT Unified Theory of Acceptance and Use of Technology
VGI Volunteered Geographic Information
1
CHAPTER I
AN INTRODUCTION TO DECISION-MAKING USING
GEOSPATIAL DATA
Abstract
This chapter provides an overview of the motivation of this dissertation, a discussion of
key industry sectors that currently use geospatial data for decision-making and a brief
history of geospatial decision-making presented through well-known cases.
Introduction
Consumer, business and government decision-makers increasingly rely on
geospatial data to make critical decisions. Recent developments, particularly the advent
of global positioning systems, expansion of advanced mobile communications networks,
prevalence of powerful mobile devices, and systems such as location-based services have
allowed individuals and organizations to collect and share vast quantities of geospatial
data. Furthermore, due to the increased access to geospatial data, as well as tools to assess
such information, many decision-makers utilize geospatial criteria in their decision-
making tasks. Modern Spatial Decision Support Systems (SDSS) have simplified such
analyses, yet there is little understanding regarding how such decision are made, what
presentation methods best facilitate geospatial decision-making, as well as what specific
factors influence decision-making performance. While multi-criterion decision-making
using geospatial data was once only practical for expert analysts, today consumers,
business and government agencies are increasingly reliant on online mapping services,
2
location-based services and SDSS to assist in their procedural, organization, and strategic
decision-making processes.
For instance, consumer tasks, such as locating the nearest bank, finding a home or
apartment, or even finding friends at an amusement park, are decision-making tasks that
are commonly aided using SDSS and associated geospatial visualization (MasterCard,
2013; Zillow, 2013; Apple, 2013). Additionally, online mapping services have become
increasingly popular with consumers, leading to over 100 million mobile device owners
accessing Google Maps each month (Gundotra, 2010). Coinciding with the increasing
usage of such mapping services, there has been an increasing public awareness of
geospatial tools and information for business and consumers in mainstream media (e.g.,
Griggs, 2013; Lavrinc, 2013; Versace, 2013).
While mobile devices have simplified everyday geospatial decision-making for
consumers, technology advances have also enhanced geospatial decision-making for
businesses, government and non-profit organizations. For instance, businesses can use
SDSS to facilitate retail site location tasks, government entities can quickly locate
citizens that will be impacted by an impending disaster, and non-profit organizations can
target individuals based on precise demographic and location criteria. More specifically,
the insurance industry can benefit from enhanced risk analytics and crisis management.
The manufacturing sector can benefit from logistic, supply chain and asset management.
The banking and finance industry can benefit from the identification of areas with high
risk potential as well as tracking consumer financial behaviors. Marketers benefit through
the use of SDSS in numerous ways as well, such as by targeting customers based on far
3
more granular locations than mailing codes alone (Hess, Rubin, & West, 2004; ESRI,
2013).
Government agencies, economic development groups and tourism authorities can
also leverage advances in SDSS. For instance, Makati City in the Philippines used SDSS
to make urban planning decisions more efficient (Africa, 2013). In the tourism industry,
SDSS has been used to perform visitor flow management, build visitor facility
inventories and assess the impacts of tourism (Chen, 2007). Economic development
groups can use SDSS to better understand specific commercial and residential markets
(ESRI, 2013).
Problem-solving using geospatial data has demonstrated its value throughout
history. A well-known example of the use of geospatial data was that of Snow (1849,
1855) who explored geospatial relationships between public wells and cholera outbreaks
to confirm that some wells were indeed contaminated and contributed to cholera
outbreaks (Brody, Rip, Vinten-Johansen, Paneth, & Rachman, 2000). Previously, Seaman
(1798) had presented an analysis of yellow fever in New York using similar thematic
maps. More recently, SDSS have been applied to better understand and mitigate various
diseases, such as malaria (Kelly, Tanner, Vallely, Clements, 2012), cancer (Rasaf,
Ramezani, Mehrazma, Rasaf, & Asadi-Lari, 2012) and diabetes (Noble, 2012). The
prevalence of the use of SDSS in medical research has even brought about a sub-domain
of healthcare research called spatial epidemiology (Elliot, Wakefield, Best, & Briggs,
2000).
Recently, geospatial data gathered and shared by large crowds using social media
has brought about exciting benefits. Such crowd-sourced data allows emergency
4
resources to be distributed more effectively following natural and man-made disasters.
For example, a United States Geological Survey (USGS) program called the Twitter
Earthquake Detector uses geospatial data and keyword filtering from Twitter data to
determine if an earthquake occurred. It has been reported that the Twitter Earthquake
Detector was able to detect a disaster within seconds, while traditional scientific alerts
took far longer to reach experts (Department of the Interior: Recovery Investments,
2012).
Academic researchers have explored the use of geospatial data for decision-
making in regard to multi-criterion decision-making, geospatial visualization, geospatial
decision-making performance, as well as group decision-making using geospatial data
(e.g.: Jankowski & Nyerges, 2001; Skupin & Fabrikant, 2003; Malczewski, 2006).
However, research in these areas has significantly lagged behind technology
developments and the increasing prevalence of SDSS, and therefore, no comprehensive
decision-making models, research constructs, measurement scales and well-established
benchmarks for geospatial decision-making exist.
Due to the increasing prevalence and access to geospatial data and an increasing
trend of individual, business and governmental decision-making being performed using
SDSS and associated information systems, it is essential that the information system (IS)
scholarship develop a more comprehensive understanding of how such decisions are
made. This current lack of understanding within the IS scholarship, as well as the
significance of geospatial decision-making, provided the primary motivation for the
development of this dissertation.
5
The next five chapters will provide initial steps toward a more comprehensive
understanding of geospatial decision-making. More specifically, Chapter II provides a
thorough literature review and research framework for geospatial decision-making
research. Chapter III presents the development of a comprehensive construct defining
individual geospatial reasoning ability, a key user-characteristic that has provided mixed
results in previous empirical studies. Chapter IV presents an extension of the Technology
Acceptance Model in the context of geospatial visualization and the use of online
mapping services. Chapter V explores decision-making within the research framework
presented in Chapter II. Together Chapter IV and Chapter V work toward extending the
external validity of the proposed geospatial reasoning ability construct. Finally, Chapter
VII presents a conclusion to the overall empirical tests of the research framework
introduced in Chapter II and defines a nomological network of the geospatial reasoning
ability construct presented in Chapter III.
6
CHAPTER II
BUSINESS DECISION-MAKING USING GEOSAPTIAL DATA: A
RESEARCH FRAMEWORK AND LITERATURE REVIEW1
Abstract
Following the brief introduction presented in Chapter I, Chapter II will present a
detailed introduction and a conceptual model for future research related to geospatial
decision-making. More specifically, Chapter II provides a thorough literature review and
framework for geospatial decision-making research.
Organizations that leverage their increasing volume of geospatial data have the
potential to enhance their strategic and organizational decisions. However, literature
describing the best techniques to make decisions using geospatial data and the best
approaches to take advantage of geospatial data’s unique visualization capabilities is very
limited. This chapter reviews the use of geovisualization and its effects on decision
performance, which is one of the many misunderstood components of decision-making
when using geospatial data. Additionally, this chapter proposes a comprehensive model
allowing researchers to better understand decision-making using geospatial data and
provides a robust foundation for future research. Finally, this chapter makes an argument
for further research of information presentation, task-characteristics, user-characteristics
and their effects on decision performance utilizing geovisualized data.
1 An early version of this chapter was submitted as a 1
st Year Paper and approved by Dawn G. Gregg in
November 2010. A subsequent version of this chapter is currently under review with Axioms.
7
Introduction
While geospatial data permeates business computing there is only a limited
understanding of how geographic information is utilized to make strategic and
organizational business decisions, as well as how to effectively visualize geographic data
for such decision-making.
The utility of database technologies, as well as that of spreadsheets, has been
taught in most business school courses for many years, so most business professionals
have had a clear understanding of utilization and benefits of such technologies. However,
as geospatial data has become more prevalent within IS, computing researchers and
business professionals have tried to better understand all aspects of decision-making
using geospatial data. One of these areas is the ability to understand decision-making
processes as they relate to the unique abilities of geovisualization, or the ability to
represent, understand and utilize geospatial data in map-like projections for decision-
making. This paper provides an analysis of current research concerning geovisualization
and decision-making performance. Additionally, this chapter responds to Vessey’s (1991)
call for additional analysis in areas of conflicting research results.
The ability to interpret geographic information and make decisions based on
geographic data is essential for business decision makers, because over 75 percent of all
business data contains geographic information (Tonkin, 1994) and 80 percent of all
business decisions involve geographic data (Mennecke, 1997). While geospatial data can
be presented utilizing traditional methods such as tables, often unique relationships
contained within geospatial data are only apparent through geovisualization (Reiterer,
Mann, Mußler, & Bleimann, 2000).
8
As organizations have collected vast amounts of geospatial or geo-referenced
data, two technologies have been developed to interpret these data in support of decision-
making. These systems are Spatial Decision Support Systems (SDSS) and Geographic
Information Systems (GIS).
While traditional DSS have been implemented successfully for production
planning, forecasting, business process reengineering and virtual shopping (Subsorn &
Singh, 2007), such systems poorly utilize geospatial data. Thus, SDSS were developed to
aid decision-making when utilizing complex geospatial data. Such technologies operate
much like DSS, but are tailored to handle the unique complexities of geospatial data.
SDSS provide capabilities to input and output geospatial data, provide analytic
capabilities unique to geospatial data, and allow complex geospatial representations to be
presented (Densham, 1991). Although IS researchers are familiar with DSS concepts,
many IS researchers are not yet familiar with key SDSS concepts such as
“georeferencing, geocoding and spatial analysis (Pick, 2004, p. 308)”.
While SDSS provide methods for geospatial decision-making, Geographic
Information Systems, commonly referred to as GIS, often allow geospatial experts to
analyze and report geographic data. GIS can be used to populate information to SDSS as
well as to perform complex geospatial analyses. While there are numerous aspects of GIS
relevant to IS researchers, this chapter examines the decision-making aspects of GIS
(Mennecke & Crossland, 1996). Such an understanding is critical for IS researchers
because the global GIS adoption rates continue to increase and research has shown that a
strong understanding of geospatial data can lead to enhanced decision-making (Pick,
2004).
9
SDSS and GIS that leverage geovisualization are provided to professionals and
consumers through a variety of sources. Prominent consumer tools for the search of
information using geovisualization include Google Maps and Bing Maps, which allow a
user to visually locate geo-referenced information, such as addresses, businesses and
even people (Google Maps, 2013; Bing Maps, 2013). Other domain-specific examples
include the capability to determine wireless signal strength at street level, automated
banking kiosk location information and interactive real estate search tools (T-Mobile,
2013; American Express, 2013; RE/MAX, 2013). However, the use of geospatial data is
not limited to only consumers and business organizations. Government agencies leverage
geospatial data for decision-making when solving large societal problems. More
specifically, immense geospatial-specific infrastructure systems have been implemented
such as the National Spatial Data Infrastructure (NSDI) in the United States and the
Infrastructure for Spatial Information in the European Community (INSPIRE) in the
European Union (Yang, Wong, Yang, & Li, 2005; Yang, Raskin, Goodchild, & Gahegan,
2010). Goals of such systems are to enable nearly every agency of a government to share
large volumes of geographic data locally, nationally and globally (Yang et al., 2005).
Early demonstration projects of the NSDI included a geospatial crime tracking system for
a metropolitan police department as well as a regional system to help communities
perform effective master planning activities (Federal Geographic Data Committee, 1999).
As more devices and technologies become networked through technologies such
as portable navigation devices, mobile computing platforms and radio-frequency
identification (RFID), more and more collected data will consist of geographic or geo-
referenced data. With the increase in the amount of geospatial data available to decision-
10
makers, it is crucial that IS professionals and researchers expand their knowledge of
geospatial systems and understand their unique characteristics, inherent potential and
limiting drawbacks. Smelcer and Carmel (1997) identified four research streams
contributing to the effectiveness of geovisualization, including information
representation, task difficulty, geographic relationship and cognitive skill. This research
chapter will expand on these findings.
Additionally, Pick (2004) points to promising research in the area of visualization,
which this research chapter addresses. Particularly, this chapter will attempt to clarify the
unique aspects that geovisualization brings to business decision-makers and will suggest
specific future research goals.
This chapter begins with a literature review emphasizing the theoretical
backgrounds of existing research, then analyzes existing research related to decision-
making utilizing geovisualization including task-characteristics, such task complexity,
collaboration, and task type and user-characteristics, such as cognitive fit, task
complexity perceptions, mental workload, goal setting, self-efficacy, spatial reasoning
ability as well as decision-making performance, all of which are identified as key
research streams relevant to the visualization of geographic data. The goal of this section
is to review relevant research in the IS, geography and psychology realms in order to
develop a comprehensive model that can be used to increase the understanding of the
impact geographic data visualization has on decision performance. Following this, a
conceptual model based on existing literature will be presented. Then, limitations of the
reviewed literature and future research suggestions will be discussed. Finally, a
conclusion is presented.
11
Literature Review
This section provides an in-depth analysis of reoccurring themes found in
literature related to task- and user-characteristics, as well as decision-making
performance.
Theoretical Background
Research reveals the importance of information-presentation, task-characteristics and
user-characteristics on decision-performance. An emphasis is placed on exploring
theories that have been suggested to explain these four themes. Specifically, literature
related to information visualization and its effects on decision-making often cites
Cognitive Fit Theory (CFT), Complexity Theory, Task Fit Theory, Image Theory as well
as research on task-technology fit, self-efficacy, motivation, goal-setting and spatial
abilities (see Table 1). Of the four research streams Smelcer and Carmel (1997)
identified, each potentially relates to an existing theory, including Task Fit (information
representation and geographic relationship), Complexity Theory (task difficulty), and
CFT (cognitive skill). This research will expand on these findings.
12
Table 1. Common Theories Related to Spatial Decision-Making.
Theory Study
Cognitive Fit Theory Vessey, 1991 (posited); Smelcer
and Carmel, 1997; Dennis and
Carte, 1998; Mennecke et al.,
2000; Speier and Morris, 2003;
Speier, 2006
Complexity Theory Smelcer & Carmel, 1997; Swink
& Speier, 1999
Task Fit/Task-Technology Fit Smelcer & Carmel, 1997;
Jarupathirun & Zahedi, 2007
Self-Efficacy Jarupathirun & Zahedi, 2007
Motivation Theory Jarupathirun & Zahedi, 2007
Goal-Setting Theory Jarupathirun & Zahedi, 2007
Image Theory Crossland, Wynne, & Perkins,
1995
The following section explores reoccurring themes found in literature related to
visualization of geospatial data. These themes include information presentation, task-
characteristics, user-characteristics and decision-performance.
Information Presentation
Numerous researchers have explored the importance of visual information
presentation on decision performance (e.g., Vessey, 1991; Smelcer & Carmel, 1997;
Dennis & Carte, 1998; Mennecke et al., 2000; Speier & Morris, 2003; Speier, 2006). For
example, in her work, Vessey posits the CFT, which suggests that there are two types of
information presentation, as well as two types of problem-solving tasks. Furthermore, it
is suggested that when the problem representation matches the problem-solving task,
higher quality decisions are made. In Vessey’s research, the objective measures of
decision time and decision accuracy, as well as interpretation accuracy, are measured as
antecedents of performance; however, it is noted that confidence in the solution also
13
could play a role. Additionally, Vessey points out that while often-analyzed tasks from
prior research utilized simple graphs and tables, actual business problems are far more
complex and not as well defined. Furthermore, prior research may have included, for
example, numbers along with graphical representations actually presenting a mix of
spatial and symbolic data.
Vessey’s (1991) CFT has been referenced as a theoretical background, extended
into other domains and validated in numerous empirical studies, such as Speier (2006)
and Smelcer and Carmel (1997). Speier presented a review of eight empirical research
papers that tested for cognitive fit. Their research discovered that all but one paper either
fully or partially supported the CFT. The author of the paper that did not support CFT,
Frownfelter-Lohrke (1998), explained that the lack of support most likely resulted due to
the complex nature of tasks involving the examination of financial statements.
Extensions of CFT include work performed by Dennis and Carte (1998) who
demonstrated that when map-based presentations are coupled with appropriate tasks,
decision processes and decision performance are influenced. Additionally, Mennecke et
al. (2000) expand on the CFT by determining the effects of subject characteristics and
problem complexity on decision efficiency and accuracy. Also, CFT has been extended
from information presentation to query interface design in order to explain how one’s
ability to understand data visualizations will influence decision outcomes (Speier &
Morris, 2003).
In addition to CFT, Task-Technology Fit has been utilized to explain the
importance of appropriate information presentation methods. For example, Ives (1982)
articulates the importance of visual information presentation. Ives states that while
14
researchers have responded to calls for additional research into data and information
visualization techniques, there is still potential for additional research into cartographic
data visualization, particularly through SDSS, GIS or other digital map-based
presentations. Specifically, Ives calls for a more in-depth understanding of how multi-
dimensional graphics could display complex information through simplified information
or charts that overlay information, both of which are technologies inherent to even basic
geovisualization systems.
Densham (1991) suggests that a SDSS user interface must be both powerful and
easy-to-use. Also, a SDSS must provide information in both graphical, or map space, and
tabular formats, or objective space, while providing the capability to move between these
representations or view these representations simultaneously to determine the most
appropriate to facilitate problem solving. However, even with multiple display options, it
is not yet understood if a decision-maker would know which of the output options
provides the best visualization method for a particular decision-making process. To
support a problem-solver who is unsure of how to select the most appropriate
visualization method, several authors have suggested the inclusion of an expert system to
provide such suggestions (Densham, 1991; Yang et al., 2005).
Additionally, relevant studies regarding the visualization of cartographic
information include Crossland et al. (1995), Smelcer and Carmel (1997), Speier and
Morris (2003) and Dennis and Carte (1998). For example, Crossland et al. performed a
study in which some participants were provided with a paper map and tabular
information while others had access to a SDSS. They were able to confirm that the
addition of a GIS-based SDSS contributed significantly toward two measures of decision-
15
making performance, decision time and decision accuracy. Speier and Morris tested the
use of text- and graphical-based interfaces to determine the effects on decision-making.
Smelcer and Carmel tested whether spatial information is best represented through
geovisualization and found that maps representing geographic relationships allowed for
faster problem solving. The authors concluded that while low difficulty tasks can be
solved quickly “regardless of representation” (p. 417), more difficult tasks “should be
represented using maps to keep problem-solving times and errors from rising rapidly” (p.
418). Dennis and Carte determined that geographically adjacent/spatial information was
best presented using spatial visualization, while non-adjacent/symbolic information tasks
were best presented using tables.
Finally, some researchers suggest that reducing the amount of information
presented to only include essential information could improve decision-making
performance (e.g., Agrawala & Stolte, 2001; Klippel, Richter, Barkowsky, & Freksa,
2005). For example, while early maps presented geospatial information with little
precision, they were still able to convey relevant information. The benefit of such
simplified maps is demonstrated by Agrawala and Stolte, who collected feedback from
over 2,000 users of a technology that emulates hand-drawn driving directions, which
often emphasize essential information while eliminating nonessential details.
Additionally, Klippel et al. suggest that modern cartographers can successfully develop
schematic maps that are simplified, yet present “cognitively adequate representations of
environmental knowledge” (p. 68). Comprehensive, yet easy-to-read, transit maps used in
large metropolitan cities demonstrate a good example of the benefit of schematization.
16
Task Characteristics
In addition to information presentation, research has shown that the specific
characteristics of the task being performed can play a vital role in decision-making
performance. Complexity Theory has been extended to demonstrate that key aspects of
geovisualization, including data aggregation, data dispersion and task complexity,
influence decision-making performance (Swink & Speier, 1999). For example,
Complexity Theory posits that as task complexity increases so too does the need for
information presentation to match problem-solving tasks. Additionally, Complexity
Theory was validated by Smelcer and Carmel’s (1997) research, which confirmed that
increased task difficulty led to decreased decision-making performance. Moreover,
Crossland and Wynne (1994) discovered that decision-making performance decreased
less significantly with the use of electronic maps, versus paper maps.
Jarupathirun and Zahedi (2001) state that, based on research by Vessey (1991),
Payne (1976), Campbell (1988) and Zigurs and Buckland (1998), tasks can be classified
into simple and complex groups based on task characteristics. Characteristics of complex
tasks include multiple information attributes, multiple alternatives to be evaluated,
multiple desired outcomes, solution scheme multiplicity, conflicting interdependence and
uncertainty.
Several empirical studies have addressed task complexity (e.g., Crossland et al.,
1995; Smelcer & Carmel, 1997; Swink & Speier, 1999; Mennecke et al., 2000; Speier &
Morris, 2003). For example, Speier and Morris (2003) discovered that decision-making
performance increased by utilizing a visual query interface when working with complex
decisions. Additionally, Swink and Speier defined task characteristics to include the
17
problem size, data aggregation and data dispersion. Their findings revealed that decision
performance, as measured by decision-quality and decision-time, was superior for smaller
problems. In the context of data aggregation, there was no effect on decision quality;
however, there was a significant effect on decision time indicating that more time was
required for disaggregated problems. Additionally, it was discovered that decision quality
for problems with greater data dispersion improved, but there was no significant effect on
decision time. Smelcer and Carmel confirmed what had been discovered in previous
research (others mentioned herein) in that more difficult tasks require additional problem
solving time. In their work, Mennecke et al. discovered that as task complexity increases,
accuracy is lowered, yet found only partial support for task efficiency being lowered.
Research conducted by Crossland et al. (1995) on the effects of SDSS on
decision-making performance included a measure of task-complexity. In their work, it
was discovered that the use of a SDSS versus data tables and paper maps significantly
improved decision-making time, while there was no significant effect on decision
accuracy. The authors pointed out that there may have been too much similarity between
the task complexity levels to ensure that decision accuracy would not be improved
through the use of an SDSS. The authors also suggested that there may be levels of
problem complexity that can only be solved through the use of an SDSS.
In their work, Albert and Golledge (1999) developed three paper and pencil tests
to assess task complexity across experience levels and gender. One of the findings was
that subjects were better at performing map overlay tasks involving ‘or’ (inclusive
disjunction) and ‘xor’ (exclusive disjunction) operators versus those utilizing ‘and’ and
‘not’ operators. The researchers also discovered that the boundary complexity of a
18
visualized entity did not affect performance, whereas the quantity of visualized entities
did. Boundary complexities can be quite varied, as for example the target radius of a
retail location may be represented by a simple circle, yet the high water boundary of a
river would be represented by a very complex boundary representing elevation, water
flow and other essential qualities.
Additional research suggests that the perception of complexity may be essential to
better understanding the effects of task characteristics on decision-making performance.
For example, Huang (2000) performed an experiment of 10 popular web-based shopping
sites and determined that increased complexity decreased the desire to explore the site,
but slightly increased the desire to purchase. Perhaps, when posited to SDSS decision-
making, an increased complexity decreases the desire to explore additional solutions
while encouraging a decision to be made quickly. This could explain some of the
variances discovered in past research particularly within the task-complexity and
decision-making performance realm. Huang’s research utilized the General Measure of
Information Rate developed by Mehrabian and Russell (1974) as a measure of the
perceived complexity.
Speier (2006) proposed a framework of complexity with four levels of
complexity; which are, in order of complexity: (1) trivial decision-making, (2) optimal
decision-making, (3) satisficing decision-making, and (4) aided decision-making.
Speier’s empirical study furthered Vessey’s (1991) CFT by comparing the outcomes of
spatial and symbolic information presentation with spatial or symbolic tasks, moderated
by task complexity, on decision performance as measured by decision quality and
accuracy. In this empirical study the subjects completed tasks that all had optimal
19
outcomes. However, findings were inconsistent with theory, as the decision time of
symbolic tasks with low complexity was reduced when using spatial presentations.
However, there are several extensions to Vessey’s work demonstrating that tables and
graphs are equal in decision performance at a low complexity and that graphs provide a
higher decision performance at high complexity.
Gill and Hicks (2006) presented a thorough list of complexity frameworks from
literature, which are presented in Table 2. Like Speier (2006), Gill and Hicks also
suggested that there are multiple classes of complexity: experienced complexity,
information processing complexity, problem space complexity, lack of structure
complexity and objective complexity.
20
Table 2. List of Complexity Frameworks (Adapted from Gill and Hicks, 2006).
Construct Type Description
Degree of Difficulty Perceived or observed difficulty
Sum of Job Characteristics Index/Job
Diagnostic Survey
Task potential to induce a state of arousal or
enrichment; measured using self-reporting
instruments such as JCI and JDS
Degree of Stimulation Task potential to induce a state of arousal or
enrichment; measured using physiological
responses
Information Load Objective measure of throughput; total
information processes or information
processed per second
Knowledge Amount of knowledge subject must possess in
order to perform task
Size Minimum theoretical size of problem space
Number of Paths Number of alternative paths that are possible
given a strategy
Degree of Task Structure Lack of strategy or structure needed to move
from initial state to goal state
Novelty of Task Uniqueness of task to subject; routine tasks
are not complex using this measure
Degree of Uncertainty Degree that the outcome of the task cannot be
predicted at initial state
Complexity of Underlying
Systems/Environment
Number of objective attributes
Function of Alternatives and Attributes Objective function of the number of
alternatives and the task attributes
Function of Task Characteristics Direct function of all possible task
characteristics
Another measure of task complexity could be manipulated through the
represented geographic relationships. For example, Smelcer and Carmel (1997) compared
spatial information used in the decision-making process presented through tables and
maps. In both the tables and the maps, common geographic relationships used in business
decision-making using spatial data were utilized. These included proximity, adjacency
and containment. Examples of proximity in the context of geographic relationships
21
include route optimization, examples of adjacency include territory assignment, and
examples of containment include site selection.
In their study of geographic containment and adjacency tasks, Dennis and Carte
(1998) discovered that when users are presented with geographic data that represents
geographic containments, tabular data presentations might lead to better decision-making,
while adjacency tasks benefit from map-based visualizations.
While most research into Complexity Theory (as pertaining to decision-making)
and Task-Technology Fit Theory has focused on individual decision performance, recent
technological innovations have led to collaborative uses of geospatial data and
information that may require these theories to be revisited through a collaborative
perspective. Geospatial data and information can lead to collaborative decision-making
through two distinct ways. First, decision-making tools utilizing geospatial information
can be used for collaborative decision-making with geographically and temporally
distributed participants. Second, through the recent phenomenon of online social
networks, geospatial information can be shared and utilized ubiquitously through vast
online communities. Each of these methods is discussed next.
A large and varied stream of research exists in the area of collaborative decision-
making utilizing geographic data. For example, grassroots groups and community
organizations have adopted Public Participation Geographic Information Systems
(PPGIS) to address the need for collaborative decision-making utilizing complex
geospatial data (Sieber, 2006). In their work, Conroy and Gordon (2004) empirically
look at the ability of a software application to increase citizen involvement in complex
policy discussions and propose that geovisualization can offer citizen participants
22
opportunities to better envision scenarios and can provide additional communication
channels to decision makers. Through the ubiquity provided by networked computing and
recent technologies, it may be possible for groups of organizations to collaborate and
form virtual organizations (Grimshaw, 2001). Jankowski and Nyerges (2001) studied the
use of GIS in a collaborative decision-making environment and discovered that decision
outcomes such as participant agreements and shared understanding could be more
effectively reached through the use of PPGIS. While there is a deep understanding of
how collaborative decision-makers can utilize geographic data, one aspect not fully
explored in current research related to geospatial information presentation and its effects
on decision-making performance was its effect on group decision-making performance.
In their framework development research, Mennecke and Crossland (1996) call for
additional exploration in the areas of GIS and its capabilities in collaborative decision-
making. Decision-making tools utilizing geospatial information can be used for
collaborative decision-making with geographically and temporally distributed
participants.
In addition to collaborative decision-making, another area of current research in
the usage of geospatial data involves how such data is utilized within online social
networks. This is especially important with the increasing use of online social networks,
because large quantities of geospatially referenced data are increasingly being shared
through such networks. Goodchild (2007) labels the geographic data that is commonly
shared through online social networks as Volunteered Geographic Information (VGI).
Some online social networks have included geographic information as a core component
in their implementations. Such availability of geographic information through online
23
social networks has even allowed researchers to map online social networks (Khalili et
al., 2009). A geographic visualization of online social networks can provide decision
makers with a geospatial representation of a virtual phenomenon. From a business
perspective, a geospatial understanding of social networks can allow strategic decision
makers to target marketing campaigns or locate retail operations in geographic areas
appropriate for their target audiences. However, the successful interpretation of VGI
from online social networks is hampered by several drawbacks. These drawbacks exist
primarily because geographic data provided by members of online social networks varies
in quality and accuracy. For example, one image may be tagged with the word “Denver”
while another is tagged with precise geographic coordinates. Additionally, as there are no
validation processes, a user can easily misidentify or intentionally provide incorrect
geospatial tagging (Khalili et al., 2010; Flanagin & Metzger, 2008).
User Characteristics
In addition to task-characteristics, researchers suggest that the characteristics of
the user also play a role in decision-performance. Such characteristics include context-
based factors, experience level, self-efficacy, cognitive workload and spatial reasoning
ability.
Several researchers have discussed the importance of context as an extension to
the research into the usability of geo-visualization tools, such as SDSS and GIS (e.g.,
Albert & Golledge, 1999; Mennecke et al., 2000; Jarupathirun & Zahedi, 2001; Slocum,
Blok, Jiang, Koussoulakou, Montello, Fuhrmann, & Hedley, 2001; Speier & Morris,
2003). Slocum et al. reported that context-based factors influence the ability to interpret
geo-visualized information. For example, “expertise, culture, sex, age, sensory
24
disabilities, education, ethnicity, physiology and anatomy, and socioeconomic status”
(Slocum et al., 2001, p. 10) are referred by the authors as influencing the ability to
interpret geospatial information. Additionally, Zipf (2002) posits that geovisualized maps
must address user contexts such as pre-existing knowledge of the area presented in the
map, physical, as well as, cognitive impairments and abilities. Specifically, an example
was demonstrated where a young child was presented with a geovisualization tool in
which abstract information was removed and a map that more closely represents reality
was presented to facilitate better interpretation. Additionally, Zipf further posits that a
user’s cultural context can influence the interpretation of the colors used in a map. For
example, in some cultures, the color green may represents parkland or forests on maps,
while in others, it represents bodies of water. In their work, Albert and Golledge tested
tasks and measured gender as a control variable. One of their conclusions was that men
performed significantly better in operations involving ‘not’ operators. An example of a
‘not’ operation in the geospatial context would be to select all neighborhoods that are
near a public school, but not near a prison. Additionally, the authors found that there were
no significant differences in test scores between subjects with GIS experience versus
those that had none. This is an essential observation as GIS and SDSS technologies are
often implemented as web-based technologies and can have users with limited
geographic visualization and information knowledge. Finally, both Zipf and Slocum et al.
point out the importance of considering sensory disabilities when developing
geovisualization technologies.
Speier and Morris (2003) discovered that task experience, database experience,
gender and computer self-efficacy were non-significant in their analysis of query
25
interface design on decision performance. Other user contexts variables, such as those
related to information learning as well as fatigue related to working through multiple
tasks, were controlled for by Swink and Speier (1999). Mennecke et al. compared
subjects with previous SDSS experience to subjects with limited SDSS experience to
determine if experience influenced efficiency and accuracy of the solutions. In their
experiments, the cognitive effort required in the decision-making process was measured
using a condensed version of the ‘Need for Cognition’ (NFC) instrument. However, their
research found only marginal support for an increase in solution accuracy and no
difference between subject groups in regard to solution efficiency. Additionally,
Mennecke et al. discovered that experience only presented significant improvement on
solution accuracy when working with paper maps. They discovered that students were
more efficient than professionals in solving geographic problems. While this may seem
surprising, it may be due to professionals incorporating multiple levels of analysis that
students, with limited experience, may not be able to draw upon.
In addition to research into the importance of user context and user experience as
related to geovisualization, other theories and constructs, particularly those from
psychology and organizational behavior, are also utilized, including self-efficacy,
motivation, goal-setting and Image Theory. For example, Jarupathirun and Zahedi (2007)
introduced a perceived performance construct consisting of decision satisfaction, SDSS
satisfaction, perceived decision quality and perceived decision efficiency. In their
findings, it was discovered that perceived decision efficiency was the greatest motivator
for goal commitment. While decision quality is likely more important than efficiency, the
26
authors proposed that there might be a perception that SDSS improves decision quality
inherently.
Additionally, Jarupathirun and Zahedi (2001) posit that based on empirical
research into the theories associated with goal-setting, users who set a higher goal level
will be motivated to expend more effort toward reaching the desired goals. Jarupathirun
and Zahedi also argue that intrinsic incentives, such as perceived effort and perceived
accuracy, can influence goal commitment levels, which are known to moderate the
effects of goal levels on performance (Hollenbeck & Klein, 1987). Finally, in an effort to
combat the lack of motivation and/or expertise, some researchers have provided financial
incentives and used experiment tasks from domains familiar to the subjects (Speier,
2006).
Crossland et al. (1995) extended Image Theory into the realm of decision-making
by proposing that the efficiencies gained through the use of electronic maps, versus paper
maps, would improve decision performance. Their study revealed that both decision-
performance, as measured through decision-accuracy and decision-time, improved with
the use of electronic maps versus paper maps at two different complexity levels.
Additionally, Jarupathirun and Zahedi (2007) discovered that self-efficacy had
strong positive influences on task-technology fit and other expected outcomes, as well as
a strong negative influence on perceived goal difficulty. It is suggested that repeated,
successful completion of tasks could improve self-efficacy, which could be accomplished
through training and learning as well as tutorials and support systems.
Another user-characteristic explored was the mental workload exhibited by
subjects performing geospatial decision-making tasks. Speier and Morris (2003)
27
measured Subjective Mental Workload (SMW) using the NASA Task Load Index
(NASA-TLX) after each completed task and discovered that when comparing visual- and
text-based interfaces, with low and high complexity decisions, the use of visual interfaces
carried a reduced SMW. Speier and Morris suggest that research into the SMW could
benefit from additional investigation. In particular, the NASA-TLX measure could use
additional validation, as the user-reported cognitive loads might not represent actual
cognitive loads that could be measured utilizing actual physiological responses.
Table 3. User-Characteristic Measurement Instruments used in Examined
Research, not including spatial ability measures (as shown in Table 4).
Study Additional User Context Instruments
Used/Cognitive Load Test Used
Speier and Morris, 2003 NASA-TLX
Mennecke et al., 2000 NFC [modified]
Jarupathirun and Zahedi, 2007 Self-Efficacy [as recommended by
Marakas et al. 1998]
Huang, 2000 General Measure of Information Rate
Finally, the importance of the spatial reasoning ability of the decision-maker
utilizing geovisualized information must be further explored, as there has been
conflicting research into the ability of spatial reasoning to aid in decision-making using
spatial data. Some research has presented no or conflicting evidence of the effects of
spatial ability on decision performance. For example, Smelcer and Carmel (1997)
discovered no statistical significance between spatial ability and the effects of
information representation, task difficulty and geographic relationships on decision
performance. The researchers speculated that due to the nature of the tasks, which did not
involve the need to navigate spatial problems, spatial visualization techniques were not
required (Smelcer & Carmel, 1997). Swink and Speier (1999) call “for more in-depth
28
investigations of visual skills related to decision-making performance” (p. 189).
Additionally, while in their early work Jarupathirun and Zahedi (2001) question whether
spatial ability has any impact on system utilization and decision-making performance,
they follow-up with a determination that spatial ability as measured through spatial
orientation ability and visualization ability had no significant effect on the perceived task-
technology fit. These findings are of value as they suggest that high spatial ability does
not influence a user’s perception of the technology. This is essential when developing a
technology for the public Internet, where it will be impractical to ensure that all users of a
technology have a prerequisite spatial ability (Jarupathirun & Zahedi, 2007).
However, other research has discovered that there are effects between spatial
ability and decision performance. For example, Swink and Speier (1999) determined that
higher spatial orientation skills produced a higher decision quality and required less
decision time; however, this finding was only significant for large problems with low
data dispersion. Additionally, Speier and Morris (2003) found that spatial reasoning
ability alone had no significant effects on decision outcomes. However, when combined
with interface design, spatial reasoning ability had a significant effect on decision-
accuracy.
Furthermore, research in other domains has identified a connection between
spatial ability and geovisualization tools. For example, Rafi, Anuar, Samad, Hayati, &
Mahadzir (2005) discusses the use of online virtual environments to facilitate the
instruction of spatial thinking skills. In their study of 98 pre-service undergraduate
students, only seven students, or about 7%, were found to have any previous spatial
experience. Rafi et al. imply that such a gap is a crucial issue and creates a hurdle for
29
students pursuing careers that require qualitative spatial reasoning. As students with no
pre-existing spatial thinking had difficulties in courses requiring spatial thinking ability, it
is likely that users lacking spatial thinking skills would have difficulties utilizing
geovisualization tools.
Additionally, students who have participated in courses that utilize
geovisualization tools, such as computerized cartography or geographic information
systems, have demonstrated improvement in their spatial thinking ability (Lee &
Bednarz, 2009). In their research, Lee and Bednarz point out that psychometric testing
designed to assess spatial abilities, such as spatial visualization and spatial orientation,
were generally focused on small-scale spatial thinking and thus not necessarily valid to
test large-scale geographic spatial abilities. However, Lee and Bednarz discovered that
recently, new spatial analysis tests have been developed which considered spatial abilities
in a geographic context (e.g., Audet & Abegg, 1996; Meyer, Butterick, Olkin, & Zack,
1999; Kerski, 2000; Olsen, 2000).
Based on these insights, the development an updated measurement instrument to
specifically determine an individual’s geospatial aptitude, which may be an essential
component of decision-making with geospatial data, is recommended. Furthermore, it is
suggested that this measurement instrument be based on the work of Lee and Bednarz
(2009) who developed and validated a spatial skills test specifically designed to
overcome shortcomings of previous spatial skills tests.
30
Table 4. Spatial Reasoning Instruments Used in Examined Research.
Study Spatial Reasoning Test Used
Smelcer & Carmel, 1997 VZ-2 (Spatial Visualization)
Albert & Golledge, 1999 Three Paper/Pencil Tests to
assess spatial ability.
Swink & Speier, 1999 S-1 (Spatial Orientation)
Speier & Morris, 2003 S-1 (Spatial Orientation)
Jarupathirun & Zahedi, 2007 VZ-2 (Spatial Visualization)
S-1 (Spatial Orientation)
Lee & Bednarz, 2009 New ‘Spatial Skills Test’
Decision-Making Performance
Another key component of geospatial decision-making is the measure of decision-
making performance. To determine decision-making performance, most researchers
utilize objective measures of decision-time and decision-accuracy as indicators of
decision-making performance, including Crossland et al. (1995), Dennis and Carte (1998)
and Speier (2006). However, in their measure of decision performance, Smelcer and
Carmel (1997) simply examined the length of time required to make a decision. Others
have proposed various additional indicators, such as decision-concept and regret (Sirola,
2003; Hung, Ku, Ting-Peng & Chang-Jen, 2007), which among other indicators, are
discussed below.
While decision-time and decision-accuracy are indicators of decision-
performance, Sirola (2003) posits that the use of an appropriate decision-analysis
methodology will undoubtedly influence decision-performance metrics and could modify
the decision maker’s perceptions of the decision-making process and result. These
decision-analysis methodologies include cost-risk comparisons, knowledge-based
systems, cumulative quality function, chained paired comparisons, decision trees,
31
decision tables, flow diagrams, pair-wise comparison, cost functions, expected utility,
information matrices, multi-criteria decision aids and logical inference/simulation.
In their vignette-based research, Speier and Morris (2003) identified decision
performance as a decision outcome, which consisted of subjective mental workload,
decision-accuracy and decision-time constructs. In their research it was discovered that
there were significant interaction effects between interface type (text/visual) and task
complexity on Subjective Mental Workload (SMW), as well as interface type and task
complexity individually.
While CFT and Complexity Theory focus on the relationships between user- and
task-characteristics and decision-performance, Goodhue’s Task-Technology Fit Theory
posits that a technology will improve task performance if the capability of the technology
matches the task to be performed (Goodhue, 1995; Goodhue & Thompson, 1995).
Smelcer and Carmel (1997) demonstrate the importance of utilizing the most appropriate
types of geographic relationships for the tasks at hand through their extension of Task-
Technology Fit Theory. Geographic relationships often found in business data include
proximity, adjacency and containment tasks. Additionally, Jarupathirun and Zahedi
(2007) synthesized research on task-technology fit with the psychology-based constructs
of goal setting and self-efficacy to further explain and determine success factors of SDSS.
The use of task-technology fit allows researchers to examine user-satisfaction, which
assess beliefs about a system and have been shown to impact adoption and intention to
use the technology (Rogers, 1983; Taylor & Todd, 1995; Karimi, Somers & Gupta,
2004).
32
In their work, Jarupathirun and Zahedi (2007) explored users’ perceived decision
quality, perceived decision performance, decision satisfaction and SDSS satisfaction and
suggest further inclusion into the Technology Acceptance Model (TAM) or the Unified
Theory of Acceptance and Use of Technology (UTAUT) models. Additionally, Dennis
and Carte (1998) discovered that when using map-based presentations, users were more
likely to utilize a perceptual decision process, while tabular data presentations induced an
analytical decision process. In their work, time and accuracy were used as measurements
of decision performance. While numerous researchers utilize perceived measures or
objective decision performance measures to determine the results of decision-making
performance, Hung et al. (2007) suggest that perceived regret should also be considered,
as many decision makers consider potential regret when making decisions. Their study
discovered significant reduction in regret for subjects who utilized a DSS over those who
did not. Other constructs and theories, particularly those from psychology and
organizational behavior, are also utilized, including research in self-efficacy, motivation,
goal-setting and Image Theory. For example, Jarupathirun and Zahedi introduced a
perceived performance construct consisting of decision satisfaction, SDSS satisfaction,
perceived decision-quality and perceived decision-efficiency. In their findings, perceived
decision-efficiency was the greatest motivator for goal commitment. While decision-
quality is likely more important than efficiency, the authors proposed that there might be
a perception that SDSS improves decision-quality inherently. These findings are
consistent with other studies in which task-characteristics have been shown to have an
impact on user satisfaction as measured through task-technology fit (Karimi et al., 2004).
33
In their study of visual-query interfaces, Speier and Morris (2003) discovered that
decision-making performance increased by utilizing a visual query interface when
working with complex decisions. In addition, Swink and Speier (1999) discovered that
moderate amounts of data dispersion required longer decision-times than tasks with low
data dispersion. In their work Albert and Golledge (1999) developed three paper and
pencil tests to test task complexity across experience levels and gender.
Based on existing literature it is evident that the most common measures of
decision-making performance are the objective measures of decision-time and decision-
accuracy. However, other research has suggested the use of perception of the decision-
making process and performance could also be utilized, particularly as user perceptions
often play a role in technology acceptance.
Conceptual Model
Based on the reviewed literature and associated theoretical frameworks, a
conceptual model of business decision-making using geospatial data is proposed. This
model consists of four distinct constructs, including information presentation, task-
characteristics, user-characteristics and decision-performance. Information presentation
was determined to be a key antecedent of decision-performance as suggested through
Vessey’s (1991) CFT. Literature has shown that different information presentation
methods may be required based on the geospatial problem being solved. Additionally,
task-characteristics have demonstrated an impact on decision-performance. Specifically,
task complexity, problem type, data dispersion, group decision-making and data quality
have been shown to define task-characteristics. In addition to information presentation
and task-characteristics, user-characteristics have also been shown to influence decision-
34
performance. Such user-characteristics include age, gender, prior experience, culture,
sensory ability, education, self-efficacy, task motivation, goal-setting, mental workload,
and geospatial reasoning ability. Finally, while decision-accuracy and decision-time were
the most common measures of decision-performance, decision-satisfaction, decision-
regret and decision-methodology could be valid measures of decision-performance.
The relationships between these constructs are presented through the following
three propositions:
Proposition P1: Information presentation impacts decision-performance.
Proposition P2: Task-characteristics impact decision-performance.
Proposition P3: User-characteristics impact decision-performance.
Figure 1 presents this conceptual model visually. This dissertation will explore
these propositions and the conceptual model through an empirical study presented in
Chapter V.
Figure 1. Proposed Research Model.
35
Discussion
The following chapters will include the design of measurement instruments and
development of laboratory experiments to test the proposed conceptual model and
propositions.
Limitations of Reviewed Literature
Four key limitations were discovered in the reviewed literature, including (1) the
choice of research subjects, (2) the selection of task types, (3) the motivation of subjects
to successfully complete the problem solving experiment and (4) a lack of experiments
testing the full conceptual model.
First, in the majority of the literature reviewed for this chapter, undergraduate
students were utilized as research subjects (see Table 5), who may not accurately
represent business decision makers who utilize geospatial data (e.g., Speier & Morris,
2003; Swink & Speier, 1999; Smelcer & Carmel, 1997). However, research conducted by
Mennecke et al. (2000) discovered that there were few differences between the results of
university students and business professionals when performing their study. Jarupathirun
and Zahedi (2007) cited Mennecke et al. (2000) in their research as a strong validation
that university students are a valid proxy for professionals. However, Jarupathirun and
Zahedi (2007) also point out that many university students are of a younger demographic
population, which might have had previous experiences with web-based SDSS such as
geovisualized search tools (e.g., restaurant and ATM locators) and online mapping tools
(e.g., Google Maps), providing them with a significant amount of prior SDSS experience.
Thus, it is recommended that future studies, which elect to use student populations,
provide a justification of the sample choice and clearly discuss limitations of
36
generalizability per the recommendations of Compeau, Marcolin, Kelley, and Higgins
(2012).
Table 5. Research Population Groups.
Study Population
Undergraduate College
Students
Crossland et al., 1995
Undergraduate Students Smelcer & Carmel, 1997
Graduate Business Students Dennis & Carte, 1998
Undergraduate Students in
Geography, Psychology and an
Introductory GIS Class
Albert & Golledge, 1999
Undergraduate Students
Previous GIS Coursework
Swink & Speier ,1999
University Students and
Professionals
Mennecke et al., 2000
Undergraduate Students Speier & Morris, 2003
Undergraduate Students Speier, 2006
Undergraduate Business
Students
Jarpathirun & Zahedi, 2007
University Students Lee & Bednarz, 2009
Second, the problem types examined by existing research have presented some
additional limitations. While some researchers chose real estate/home finding as their
task method (e.g., Speier & Morris, 2003), others chose more domain specific tasks. This
is a concern as, for example, the task of locating an automated banking kiosk would
reflect a fundamentally different problem than determining properties that may be
impacted by a natural disaster. It is suggested that future research carefully select
problem types to facilitate an improved comparison of research results.
Third, the motivation for completing the research tasks accurately can be
questioned. To address this issue, some researchers provided monetary incentives to
participants plus additional monetary incentives for higher decision performance (Hung
et al., 2007). Furthermore, simulated experiments may not adequately reveal how
37
individuals make ‘real world’ decisions, as the motivations might be different.
Additionally, group decision-making may have different motivations than individual
decision-making.
Finally, few studies measured task- and user-characteristics simultaneously with
information presentation to determine moderating impacts of each construct. Thus, it is
suggested that the entire conceptual model be tested empirically to determine the effects
of each antecedent construct on decision-performance.
In addition to the addressing these limitations, numerous future research
opportunities have presented themselves in the course of this literature review.
Future Research
It is suggested that additional work be performed to test the proposed
comprehensive model utilizing varying task and user characteristics which could help
identify further moderating and mediating variables.
For example, Swink and Speier (1999) suggest that complexity levels could be
increased to determine if their findings still hold true. Albert and Golledge (1999) call for
more research into specific tasks and how groups of individuals are able to make-
decisions using geovisualization.
In their work, Jarupathirun and Zahedi (2007) measured effects of perceptual
constructs including perceived efficiency and perceived accuracy; however, a comparison
to objective measures was not made and the authors suggested such an experiment as
potential future research. Additionally, the authors suggested that user’s perceptions over
time could be measured to develop a more comprehensive understanding. Crossland et al.
38
(1995) suggest that future research assess “decision-maker confidence, user process
satisfaction, and individual level of motivation” (p. 232).
The surveyed literature suggested numerous future research possibilities within
the four research themes presented within this paper. These themes include information
presentation, task-characteristics and user-characteristics and their effects on decision
performance using geospatial data. Of these, the theme of user characteristics seems most
promising, particularly in regard to a deeper understanding of the importance of
geospatial ability. Some general research questions, based on the conceptual model
include:
Which geovisualization techniques improve decision-performance?
Which specific user-characteristics impact decision-performance?
Can specific geovisualization techniques overcome user-characteristics
that negatively impact decision-performance?
Which specific task-characteristics impact decision-performance?
Can specific geovisualization techniques overcome task-characteristics
that negatively impact decision-performance?
Potential research questions related to geospatial ability include:
Do current spatial ability measurement instruments properly identify the
ability to analyze and interpret geospatial data as would be defined in a
geospatial ability construct? If not, what key antecedents can measure
geospatial ability?
Does geospatial ability influence decision-making performance? And if so,
can inconsistencies in past empirical studies be explained?
39
Do actual geospatial ability and perceived geospatial ability differ?
To answer these questions, it is proposed that a geospatial ability construct be
developed along with a comprehensive measurement instrument. Such a construct would
allow researchers to more easily validate research results which could provide business
leaders with a better understanding of the need of geospatial ability in the workforce.
Furthermore, an understanding of the importance of geospatial ability could lead to
refined geovisualization tools which may overcome potential gaps in a user’s geospatial
ability. Additionally, it is suggested that future geospatial decision-making research
emphasize the importance of including measures for each of the three antecedent
constructs (information presentation, task-characteristics, user-characteristics) to
determine their combined effects on decision-performance. Future research into decision-
making using geospatial data should continue to validate existing theory as well as
provide business decision-makers with sound best practices and tools for decision-
making. Furthermore, an understanding of the importance of geospatial decision-making
could lead design science researchers to develop refined geovisualization tools which
may overcome potential negative task- and user-characteristics a user’s geospatial ability.
This dissertation will include the development of a geospatial reasoning ability
construct, measures of such a construct, as well as empirical research to determine the
potential benefits of such a construct.
Conclusion
As organizations collect large amounts of geospatial data, there is a need to
effectively utilize the collected data to make strategic and organizational decisions.
However, literature describing the best techniques to make decisions using geospatial
40
data as well as the best approaches for geovisualization is limited. This literature review
revealed that existing research can provide a strong foundation for future exploration of
how business decision-making using geospatial data occurs. Additionally, a conceptual
model for the study of effects of geovisualization on decision-performance is presented
and defined through existing theory. Along with the conceptual model, numerous
applicable research methods, existing constructs, potential limitations, validity concerns
and potential future research questions were presented.
While business school curriculums often include basic computing courses, as well
as courses designed to convey understanding of database and spreadsheet tools, there has
been neglect in incorporating even a basic understanding of the importance of geospatial
data as well as the unique tools to interpret such data. While in the last ten years some
business schools have recognized the need for the inclusion of GIS by offering a required
course, an elective course or even a degree emphasis or research center (Pick, 2004),
following the insights represented in this paper, it is recommended that business
curriculums include problem-solving exercises utilizing geospatial data.
Based on the numerous contradictions in past research, particularly pertaining to
task complexity, the development of common framework for determining task
complexity levels in problem-solving scenarios is recommended. Such a framework
could then be used to benchmark research and provide researchers with the ability to
more easily compare research results. Additionally, based on discrepancies of previous
research into geospatial ability, the importance of spatial ability in problem solving using
geovisualized information must be better understood in order to ensure that businesses
and individuals are able to make better decision using geographic data. This research will
41
allow information system practitioners to effectively utilize geovisualization tools to
organize and present large quantities of geospatial data.
This chapter reviewed the use of geospatial visualization and its effects on
decision performance, which is one of the many components of decision-making using
geospatial data. Additionally, this chapter proposed a comprehensive model allowing
researchers to better understand decision-making using geospatial data and provided a
robust foundation for future research. Finally, this chapter made an argument for further
research into task- and user-characteristics and their effects on decision-performance
when utilizing geospatial data.
42
CHAPTER III
GEOSPATIAL REASONING ABILITY: CONSTRUCT AND
SUBSTRATA DEFINITION, MEASUREMENT AND VALIDATION2
Abstract
The preceding chapter suggested several potential research question related to
geospatial reasoning ability. The first of these questions relates to determining how
geospatial reasoning ability can be measured, as well as what the key antecedents of
geospatial reasoning ability are. Thus, Chapter III presents the development of a
comprehensive construct defining individual geospatial reasoning ability, a key user-
characteristic that has provided mixed results in previous empirical studies.
Today’s organizations often gather large quantities of geographic and geospatially
referenced data to support business decision-making. Prior research investigating the
significance of user-characteristics on decision-performance, when working with
geospatial data, has presented conflicting results. This is particularly true in regard to the
impact of geospatial reasoning ability on the ability to perform efficient and effective
decisions making. As the amount of geographic- and geospatially-referenced data grows,
it is essential to develop a comprehensive understanding of how user characteristics, such
as geospatial reasoning ability, influence decision-performance. Furthermore, such an
understanding is an essential component within the human-computer interaction domain.
This research introduces two new constructs, Geospatial Reasoning Ability and
2 An earlier version of this chapter was submitted as a Comprehensive Examination Paper and approved in
September 2012. A subsequent version of this chapter was presented at the Americas Conference on
Information Systems (AMCIS) 2011 and is currently under review with Decision Support Systems.
43
Geospatial Schematization, and presents validated measurement scales to accompany
these constructs.
Introduction
Geospatial data permeates business computing. Many business decisions, from
transport and logistics to marketing and product development, now rely on geospatial
data. This highlights the need for an improved understanding of ways to best utilize
geographic data when making such decisions. Scholars have stated that over 75 percent
of all business data contains geographic information (Tonkin, 1994) and that 80 percent
of all business decisions utilize geographic data (Mennecke, 1997). It is very likely that
these percentages are now even greater due to the prevalence of devices that can collect
geo-referenced data, particularly GPS-enabled mobile computing platforms, such as
smartphones and tablets.
Researchers have found that a clear understanding of such geospatial data can
lead to improved business decision-making (Pick, 2004). Specialized tools have been
developed to help businesses utilize the immense body of geospatial knowledge that is
continuously being collected. These tools include Spatial Decision Support Systems
(SDSS) and Geographic Information Systems (GIS).
SDSS function much like Decision Support Systems (DSS), but are designed to
enable business decision-makers to better understand the complexities of geospatial data.
GIS allow experts to analyze and report geographic data. As various GIS and SDSS tools
evolve, many of their functionalities and capabilities are converging. Of additional
interest is that these tools have evolved to allow non-geospatial experts to easily perform
decision-making functions using geospatial data. For example, many banks provide
44
online and mobile tools allowing consumers to locate automated banking kiosks with
particular criteria, such as hours of operation and types of cards accepted.
As the volume of geographic and geographically-referenced data grows and
organizations rely on geovisualization to present such data to business decision-makers, it
is imperative that all aspects of the decision-making process are fully understood in order
to help researchers and practitioners develop optimal systems. This chapter provides the
first step toward improving our understanding of how geospatial reasoning ability
impacts the capability of business decision-makers to use geovisualization when making
decisions.
There are several research questions related to geovisualization, decision
performance and geospatial reasoning ability. First, can geospatial reasoning ability be
measured practically? Second, does geospatial reasoning ability moderate decision-
making performance? Third, can specific geovisualization techniques allow a decision
maker without strong geospatial reasoning ability to make sound decisions? The first of
these questions is addressed in this chapter.
This chapter begins with a literature review emphasizing the theoretical
background and the current understanding of user characteristics, particularly spatial
reasoning, on geospatial decision performance. Then, a section describing the scale
development procedure is presented. This is followed by a formal definition of the
proposed construct and substrata, along with a measurement model. Next, an exploratory
study is presented and discussed. Finally, limitations, future research suggestions and a
conclusion are presented.
45
Literature Review
This section provides an analysis of reoccurring topics found in literature related
to geovisualization and decision performance, along with a review of literature which
explores the effects of user-characteristics on decision-making performance, with an
emphasis on research related to spatial reasoning. As business leaders are faced with
tasks such as the interpretation of large quantities of business data containing geospatial
information, it is crucial for researchers to understand how such decisions are made.
Particularly, what presentation formats are most appropriate, which individual
characteristics form stronger decision-makers, and what specific decision tasks can be
made utilizing geospatial data? Applying an existing theoretical lens to this problem will
allow researchers to draw upon established research in an attempt to answer these
questions.
Cognitive Fit Theory
Numerous researchers have explored the importance of visual information
presentation. For example, Vessey (1991) posits the CFT, which suggests that when
information presentation matches the problem-solving task, higher quality decisions are
made. The CFT has been referenced as the theoretical background, extended into other
domains and validated in numerous empirical studies, such as Mennecke et al. (2000),
Smelcer and Carmel (1997) and Speier and Morris (2003). Extensions of the CFT include
work performed by Dennis and Carte (1998), which demonstrated that when map-based
presentations are coupled with appropriate tasks, decision processes and decision
performance are influenced. Additionally, Mennecke et al. expand on the CFT by
determining the effects of subject characteristics and problem complexity on decision
46
efficiency and accuracy. Speier (2006) presented a review of eight empirical research
papers that tested for cognitive fit. This research discovered that all but one paper
(Frownfelter-Lohrke, 1998) either fully or partially supported the CFT.
Geovisualization and Decision-Performance
While geospatial data can be presented utilizing traditional methods such as
tables, unique relationships contained within geospatial data are often only apparent
though geovisualization (Reiterer et al., 2000). Geovisualization, or the visualization of
geospatial or geospatially-referenced data, and its effect on decision performance has
been explored in several key works, including Crossland et al. (1995), Smelcer and
Carmel (1997), Speier and Morris (2003) and Dennis and Carte (1998).
Research has revealed that decision performance was improved through the use of
GIS-based SDSS (Crossland et al., 1995), graphical-based interfaces for geospatial data
(Speier & Morris, 2003), geovisualization for tasks with a high task-complexity (Smelcer
& Carmel, 1997) and for presenting adjacent information (Dennis & Carte, 1998).
Comparisons were made with paper maps (Crossland et al., 1995), text-based interfaces
to geospatial data (Speier & Morris, 2003) and tabular presentation (Dennis & Carte,
1998).
Research that examines information presentation and its effects on decision
performance utilizes numerous measurement indicators; however, objective measures,
such as decision-time and decision-accuracy, are the most common (Crossland et al.,
1995; Dennis & Carte, 1998; Smelcer & Carmel, 1997; Speier, 2006). Other indicators
have included perceptions of decision-performance and even decision-regret (Hung et al.,
2007; Sirola, 2003). Researchers exploring the relationship of geovisualization on
47
decision-performance also often include task characteristics, user characteristics, or both,
in their work.
User-Characteristics
There is growing research into the effects of user characteristics, particularly
context-based (Albert & Golledge, 1999; Slocum et al., 2001; Zipf, 2002) and cognitive-
based (Mennecke et al., 2000; Speier & Morris, 2003). Furthermore, user-characteristics
explored in research have included gender (Albert & Golledge, 1999), previous
experience with SDSS tools (Mennecke et al., 2000), self-efficacy and motivation
(Jarupathirun & Zahedi, 2001, 2007).
While the existing work on user characteristics has revealed important insights,
research into the importance of spatial reasoning ability has produced conflicting results
(Smelcer & Carmel, 1997; Swink & Speier, 1999; Speier & Morris, 2003; Rafi et al.,
2005; Jarupathirun & Zahedi, 2007; Lee & Bednarz, 2009). For example, Smelcer and
Carmel discovered no significant relationships between spatial ability and decision-
performance. However, the researchers did point out that this might be because the task-
characteristics may not have required the use of spatial ability from the research subjects.
Additionally, Swink and Speier noted inconsistencies in prior research measuring spatial
ability and called for additional research. Finally, Jarupathirun et al. revealed that spatial
ability, as measured through spatial orientation ability and visualization ability, had no
effect on decision performance.
In contrast, another corpus of research had found relationships between spatial
ability and decision-performance. For instance, Swink and Speier (1999) revealed that
experiments utilizing task-characteristics involving large problems with low data
48
dispersion and user-characteristics of high spatial orientation improved decision
performance. Rafi et al. (2005) stated that students who participated in virtual
environments had improved spatial abilities, which could help overcome difficulty in
courses that require spatial skills, such as engineering graphics or cartography. Whitney,
Batinov, Miller, Nusser and Ashenfelter (2011) determined that address-verification
performance was significantly associated with spatial ability. Alternately, Lee and
Bednarz (2009) discovered that students who completed courses utilizing geospatial tools
gained greater geospatial reasoning ability as measured by a geospatial ability
measurement exam developed by the researchers. Rusch, Nusser, Miller, Batinov and
Whitney (2012) discovered that spatial ability has an effect on decision-performance
when testing for three dimensions of spatial ability: spatial visualization, logical
reasoning and perspective taking. These conflicting results may be due to the nature of
the spatial reasoning tests used (see Table 6).
49
Table 6. Spatial Reasoning Instruments Used in Examined Research
Study Spatial Reasoning Test
Used
Effect of Spatial Ability on
Outcome
Smelcer et al., 1997 VZ-2 (Spatial
Visualization)
Effect
Albert et al., 1999 Three Paper/Pencil Tests Partial Effect
Swink et al., 1999 S-1 (Spatial Orientation) Effect
Speier et al., 2003 S-1 (Spatial Orientation) Effect
Jarupathirun et al.,
2007
VZ-2 (Spatial
Visualization)
S-1 (Spatial Orientation)
No Effect
Lee et al., 2009 ‘Spatial Skills Test’ Effect
Whitney et al., 2011 VZ-2 (Spatial
Visualization)
MV-2 Visual Memory
PT (Perspective Taking)
Effect
Rusch et al., 2012 VZ-2 (Spatial
Visualization)
Logical Reasoning
PT (Perspective Taking)
Effect
Lee and Bednarz (2009) emphasize that most of the spatial ability tests utilized in
research involving geospatial ability are based on ‘table top’ psychometric exams, such
as those that measure spatial orientation and spatial visualization. Indeed, many of the
reviewed studies utilized various components of popular spatial reasoning tests (see
Table 6). To overcome this limitation, Lee and Bednarz (2009) developed a test to
measure spatial reasoning within a geographic context.
Task-Characteristics
Research has shown that characteristics of the task being performed can play a
vital role in decision-making performance. For example, Complexity Theory (Campbell,
1988) posits that as task complexity increases so too does the need for information
50
presentation to match problem-solving tasks. Complexity Theory has been extended to
demonstrate that key aspects of geovisualization, including data aggregation, data
dispersion and task complexity, influence decision-making performance (Swink and
Speier, 1999). Additionally, Complexity Theory was validated by Smelcer and Carmel’s
(1997) research, which confirmed that increased task difficulty led to decreased decision-
performance. Moreover, Crossland and Wynne (1994) discovered that decision-making
performance decreased less significantly with the use of electronic maps versus paper
maps.
Jarupathirun and Zahedi (2001) state that, based on research by Vessey (1991),
Payne (1976), Campbell (1988) and Zigurs and Buckland (1998), tasks can be classified
into two groups, simple and complex, based on task-characteristics. Characteristics of
complex tasks include multiple information attributes, multiple alternatives to be
evaluated, multiple desired outcomes, solution scheme multiplicity, conflicting
interdependence and uncertainty.
Several empirical studies have addressed task complexity. For example, Speier
and Morris (2003) discovered that decision-making performance increased by utilizing a
visual query interface when working with complex decisions. Additionally, Swink and
Speier (1999) defined task characteristics to include the problem size, data aggregation
and data dispersion. Their findings indicated that decision-performance, as measured by
decision-quality and decision-time, was superior for smaller problems. In the context of
data aggregation, there was no effect on decision quality; however, there was a
significant effect on decision-time, indicating that more time was required for
disaggregated problems. Additionally, it was discovered that while decision-quality for
51
problems with high data dispersion improved, there was no significant effect on decision-
time. Smelcer and Carmel (1997) confirmed what had been revealed by previous research
(others mentioned herein), that more difficult tasks require greater decision-time. In their
work, Mennecke et al. (2000) discovered that as task complexity increases, accuracy is
lowered, yet found only partial support for task efficiency being lowered.
Scale Development Procedure
As many of the reviewed research projects explored different dimensions of
spatial ability, and most did not measure spatial reasoning with regards to geographic
context, this could partially explain the contradictory results found in these studies. This
finding suggests a need for improved measures of spatial ability which is context-
sensitive to both geographic-scale (task aware) and business decision-making (user
aware.) Thus, a new construct and measurement scale, Geospatial Reasoning Ability
(GRA), are introduced in the following sections as a measure of an individual’s cognitive
geospatial reasoning ability within the context of performing business decision-making.
This research utilized the construct development and validation procedures
recommended by MacKenzie, Podsakoff, & Podsakoff (2011) to define and measure
geospatial reasoning ability and to develop an appropriate measurement scale (Figure 2).
In their work, MacKenzie et al. present a ten-step scale development procedure. These
steps are classified into six categories: conceptualization, development of measures,
model specification, scale evaluation and refinement, validation, and norm development.
This chapter completes the first seven steps in the scale development process.
Completing these steps allows for an initial GRA scale to be developed and validated.
Future research studies will be necessary to validate this scale in other contexts and to
52
develop norms for the scale. This is consistent with other initial scale development
studies (e.g. Balog, 2011; Brusset, 2012).
Figure 2. Overview of Scale Development Procedure from MacKenzie et al. (2011).
53
Construct Conceptualization (Step 1)
MacKenzie et al. (2011) recommend that the first step of the scale development
procedure is to complete four factors of construct conceptualization. These four factors
include: (1) establishing how the construct has been used in prior research or by
practitioners, (2) identifying the properties of the construct as well as the entities to which
the construct applies, (3) specifying the conceptual theme of the construct and (4)
defining the construct in clear, concise and unambiguous terms.
Factor One: Examination of Prior Research
Factor one of construct conceptualization, as suggested by MacKenzie et al.
(2011), is to conduct a comprehensive literature review to determine how geospatial
reasoning and similar concepts have been defined and measured. The majority of these
findings were described in the literature review section above.
Prior research has shown conflicting results when the relationship between spatial
ability and decision performance was assessed (e.g. Jarupathirun & Zahedi, 2007; Rafi et
al., 2005; Smelcer & Carmel, 1997; Speier & Morris, 2003; Swink & Speier, 1999), as
summarized in Table 6. The majority of these studies have had two major limitations: (1)
they only assessed certain dimensions of spatial ability and (2) they do not specifically
address the geographic context. Furthermore, existing measurement tools are often
complicated to administer and may not provide comparative results as only certain
dimensions are measured. For example, some of these tests require prior GIS knowledge
in order to be completed successfully.
To address these concerns, Lee and Bednarz (2009) developed a new spatial skill
test. Their test was designed to overcome the aforementioned limitations, by measuring
54
multiple dimensions of spatial thinking and by placing these questions in a geospatial
context. Their 30-minute test consisted of seven sets of question items, including
performance tasks and multiple-choice questions. While this assessment addressed many
of the initial concerns, it too had some limitations. For example, the scale was designed to
evaluate, and be tested on, students enrolled in a geography department and thus the
questions specifically refer to many common GIS tasks, such as site selection and
topography. Concepts and terms such as these are ones that a business decision-maker
may not be very familiar with, requiring training in order to clarify the test tasks to
subjects outside the geography domain. Furthermore, a post-study questionnaire revealed
that the subjects perceived many of the questions as too simple.
We address these limitations through the development of a measurement scale
that (1) addresses known dimensions of geospatial reasoning, (2) emphasizes the
geographic context of the cognitive ability measured and (3) assesses GRA in subjects
not necessarily familiar with advanced cartographic and geographic terms. Doing so will
address the major limitations of the existing measurement tools.
The limitations found in the current body of research are the primary motivation
for the development of a new construct and measurement tool. Such a scale could provide
a more accurate measure as it would be sensitive to geographic-scale (task aware) and
business decision-making (user aware). Furthermore, such a scale would allow future
researchers to more easily incorporate geospatial reasoning ability in their work. Thus, a
new construct, Geospatial Reasoning Ability (GRA) of business decision makers, is
introduced as a measure of an individual’s geospatial reasoning ability within the context
55
of performing business decision-making, along with an easy-to-use questionnaire-based
measurement scale.
Factor Two: Identification of Construct Properties and Entities
MacKenzie et al. (2011) suggest that the second factor of the construct
conceptualization step should include the identification of the construct properties and
identification of the entities to which it applies. As previous studies measuring spatial
ability were focused on individual decision-making, the proposed construct is also
applied to an individual entity. However, since teams normally make many business
decisions, the impact of GRA on group decisions should be examined in the future.
While spatial reasoning has been measured using cognitive tests (Speier &
Morris, 2003) and psychometric test (Jarupathirun & Zahedi, 2001, 2007), this research
utilizes psychometric measures to estimate GRA by measuring self-perceived traits.
These self-perceived traits are referred to as substrata in this paper. Cognitive tests are
often used in laboratory experiments where activities can be more easily timed and
controlled, while psychometric tests can often be administered in a variety of settings. As
one of the goals of this research was to develop an easy-to-administer measure of GRA,
psychometric measures were utilized.
Thus, in the context of this research, the proposed GRA construct specifies a type
of cognition that applies to an individual person, and is measured utilizing self-perceived
substrata.
56
Factor Three: Specification of the Conceptual Theme
The next factor of the construct conceptualization step, as suggested by
MacKenzie et al. (2011), is the specification of the conceptual theme. We propose that
the GRA construct specifies a cognitive ability that consists of multiple dimensions based
on self-perception. While prior research has focused on specific psychometric
components of spatial ability (e.g., Smelcer & Carmel, 1997, Swink & Speier, 1999), this
study attempts to establish a model encompassing all known and measurable substrata of
GRA as identified through a comprehensive literature review.
MacKenzie et al. (2011) also suggest specifying if a construct changes or remains
stable over time. We suspect GRA to remain stable over time; however, specialized
training or careers that utilize geospatial thinking may influence a person’s GRA. For
example, it has been shown by Rafi et al. (2005) that spatial intelligence can be improved
through specific training.
Factor Four: Definition of the Construct and Substrata
As suggested by MacKenzie et al. (2011), the final factor is to unambiguously
define the construct. GRA can be defined as a business decision maker’s cognitive ability
to assess and utilize geospatial information in order to make sound business decisions. To
measure this construct, the various substrata of GRA are identified and defined next.
In order to identify and define the substrata of the GRA construct, a literature
review of published works, that specifically explored spatial and geospatial reasoning
ability was conducted. During this process, scholarly articles were reviewed for key
semantic content (e.g. spatial ability, spatial thinking, spatial reasoning, spatial skills,
geospatial ability, geovisualization, maps, and geographic information systems) by an
57
iterative approach using the search capabilities of EBSCO Publishing’s Business Source
Premier and Thomson Reuters’ Web of Science. This review revealed several potential
substrata of the GRA construct. To validate the results of this step, two academic and two
industry experts were asked to validate each of the proposed substrata and their
definitions for agreement. The two industry experts, who work with and develop
geovisualization systems, had over 20 years of combined experience. The academic
experts perform research on behavioral/cognitive topics in the information systems
scholarship. Each of these experts suggested only minor updates to the definitions and
stated an overall agreement with the proposed substrata. The final substrata and
definitions are presented in Table 7.
Table 7. Proposed Substrata of GRA.
Proposed Substrata Definition Developed From
Self-Perceived Geospatial
Orientation and
Navigation (SPGON)
The self-perceived ability to
determine one’s position and
direction in geographic
space.
Kozlowski & Bryant,
1977; Swink & Speier,
1999; Cherney, Brabec
and Runco, 2008; etc.
Self-Perceived Geospatial
Memorization and Recall
(SPGMR)
The self-perceived ability to
commit geographic concepts
to memory and the ability to
reconstruct these concepts.
Lei, Kao, Lin & Sun,
2009; Miyake, Friedman,
Rettinger, Shah &
Hegarty, 2001; etc.
Self-Perceived Geospatial
Visualization (SPGV)
The self-perceived ability to
form mental images of
geographic space.
Eisenberg & McGinty,
1977; Velez, Silver &
Tremaine, 2005; etc.
A second construct, which we have labeled Self Perceived Geospatial
Schematization (SPGS), has been identified in literature as important for decision-making
using geospatial data, yet, no research was found that specifically identified
schematization ability as an indicator of spatial reasoning ability. However, several
studies indicated its importance for communicating geospatial information (e.g., Klippel
58
et al., 2005). The definition of SPGS is presented in Table 8. Based on the potential of
SPGS to be a potential substratum of GRA, it was tested as both an independent construct
as well as a potential substratum of GRA.
Table 8. Definition of SPGS.
Construct Definition Developed From
Self-Perceived Geospatial
Schematization (SPGS)
The self-perceived ability to
reduce complexity of
geographic elements by
converting to a schema or
outline.
Agrawala & Stolte, 2001;
Klippel et al., 2005; etc.
Following the completion of the conceptualization step of the scale development
procedure outlined by MacKenzie et al. (2011) the two-step process of measurement item
generation is performed.
Measurement Item Generation (Step 2)
While existing measurement tools often explore spatial abilities, there are distinct
differences between the general spatial abilities measured by these tools and spatial
reasoning abilities in a geographic context. Furthermore, the existing spatial ability
measurement tools that were reviewed only address some components of spatial ability,
such as visualization, not the proposed spectrum of GRA. Additionally, many of these
existing measurement tools often require the subjects to complete physical or visual
tasks, making the tests difficult to administer. Such measurement tools usually require
subjects to be physically present in a laboratory setting. The goal of this research was to
develop a set of measurement items that correctly measure the various dimensions of
GRA and that are easy to administer. Given that measurement items for each of these
substrata do not exist in literature, the author developed each item. The measurement
59
items were developed utilizing semantic terminology derived from the substrata
definitions and existing literature. The items were further refined through the opinions of
the experts mentioned above.
The following tables (Table 9-11) present the initial measurement items of the
proposed substrata of the GRA construct, which include self-perceived geospatial
orientation and navigation (SPGON), self-perceived geospatial memorization and recall
(SPGMR) and self-perceived geospatial visualization (SPGV).
The SPGON substratum is defined as ‘the self-perceived ability to determine
one’s position and direction in geographic space.’ The following initial measurement
items (see Table 9) are expected to measure the SPGON substratum.
60
Table 9. Initial Self-Perceived Geospatial Orientation and Navigation Measurement
Items.
Item ID Item
SPGON1 In most circumstances, I feel that I could quickly determine where I
am based on my surroundings.
SPGON2 I can usually determine the cardinal directions by looking at the sky.
SPGON3 I rarely get lost.
SPGON4 Examining my surroundings allows me to easily orient myself.
SPGON5 At any given point, I usually know where north, south, east and west
lie.
SPGON6 I find it difficult to orient myself in a new place.
SPGON7 Knowing my current location is rarely important to me.
SPGON8 I feel that I can easily orientate myself in a new place.
SPGON9 Using a compass is easy.
SPGON10 I rarely consider myself lost.
SPGON11 I have a great sense of direction.
SPGON12 I could easily navigate a course of ‘500 meters north and 500 meters
east’ using a magnetic compass.
SPGON13 Knowing my current location is essential to determine where I am
going.
SPGON14 When driving or walking home, I am likely to choose a route that I
have never taken.
Author Developed, Source for concepts: Kozlowski & Bryant , 1977; Swink &
Speier, 1999; Cherney et al., 2008; Meilinger & Knauff, 2008
The SPGMR substratum is defined as ‘the self-perceived ability to commit
geographic concepts to memory and the ability to reconstruct these concepts.’ The initial
SPGMR measurement items, shown in Table 10, are expected to measure the SPGMR
substratum.
61
Table 10. Initial Self-Perceived Geospatial Memorization and Recall Measurement
Items.
Item ID Item
SPGMR1 I am good at giving driving directions from memory.
SPGMR2 I am good at giving walking directions from memory.
SPGMR3 I can usually remember a new route after I have traveled it only once.
SPGMR4 I don’t enjoy giving directions, as I have trouble recalling the details
needed.
SPGMR5 I don’t remember routes.
SPGMR6 When revisiting a place I don’t frequent often, I usually can
remember how to get around.
SPGMR7 After studying a map, I can often follow the route without needing to
look back at the map.
SPGMR8 When someone gives me good verbal directions, I can usually get to
my destination without asking for additional directions.
SPGMR9 After being a shown a map, it would be easy for me to recreate a
similar map to memory.
SPGMR10 After seeing a city map once, I am usually able to commit key
landmarks and their locations to memory.
SPGMR11 It would be easier to memorize a tabular list of states, instead of a
map showing the states.
SPGMR12 It would be easier to memorize capital cities from a map, instead of a
table.
SPGMR13 The best way to memorize the layout of a college campus is to study
a map.
SPGMR14 I could draw an approximate outline of my home country from
memory.
Author Developed, Source for concepts: Lei et al., 2009; Miyake et al., 2001
The SPGV substratum is defined as ‘the self-perceived ability to form mental
images of geographic space.’ The initial SPGV measurement items, shown in Table 11,
are expected to measure the SPGV substratum.
62
Table 11. Initial Self-Perceived Geospatial Visualization Measurement Items.
Item ID Item
SPGV1 It is easy for me to visualize a place I have visited.
SPGV2 I find it difficult to visualize a place I have visited.
SPGV3 I can visualize a place from information that is provided by a map
without having been there.
SPGV4 When someone describes a place, I form a mental image of what it
looks like.
SPGV5 I can visualize geographic locations.
SPGV6 When someone describes a place, I have a difficult time visualizing
what it looks like.
SPGV7 When viewing an aerial photograph, I often visualize what the area
looks like on the ground.
SPGV8 I can visualize a place from an aerial photograph.
SPGV9 I can visualize a place from a verbal description.
SPGV10 I can visualize a place from a map.
SPGV11 I can visualize what a future building might look like on an empty
lot.
SPGV12 When viewing a map, I often visualize what the area looks like on
the ground.
SPGV13 While reading written walking directions, I often form a mental
image of the walk.
SPGV14 Generally I prefer to memorize the mental images of a walk or drive,
versus the written directions.
Author Developed, Source for concepts: Eisenberg & McGinty, 1977; Velez et
al., 2005
Table 12 presents the initial measurement items of the proposed self-perceived
geospatial schematization (SPGS) construct. The SPGS construct is defined as ‘the self-
perceived ability to reduce complexity of geographic elements by converting to a schema
or outline.’
63
Table 12. Initial Self-Perceived Geospatial Schematization Measurement Items.
Item ID Item
SPGS1 I prefer maps that display key information clearly, such as transit
maps.
SPGS2 I prefer maps that include full-color aerial photography.
SPGS3 I prefer simple, sketch-like maps.
SPGS4 When looking at subway or transit maps, I can usually quickly find
the routes I need to take in order to reach my destination.
SPGS5 When giving directions it is easy for me to decide what is important
enough to include and what to exclude.
SPGS6 I prefer maps that only provide key information, even if they are not
to scale.
SPGS7 I prefer maps that only provide information necessary to accomplish
my tasks.
SPGS8 I prefer written walking directions that only include the most
essential navigational elements.
SPGS9 I prefer maps that show only the most essential information.
SPGS10 I am better at interpreting maps that only provide necessary
information.
SPGS11 It is easy for me to ignore irrelevant information on a map and to
focus only on necessary information.
SPGS12 I prefer verbal driving directions that only include the most essential
information to reach my destination.
SPGS13 I like simple, clear maps.
SPGS14 I prefer highly detailed maps that show more than just the basic
information.
Author Developed, Source for concepts: Agrawala & Stolte, 2001; Klippel et
al., 2005
Content Validity (Step 3)
Following the initial generation of items, MacKenzie et al. (2011) recommend
establishing content validity. In order to establish content validity, a categorization and
prioritization exercise was conducted. Ten business decision makers, who have worked
as functional or project managers for ten years or more, were selected to perform this
64
exercise. Due to their professional experience, these business decision-makers were ideal
candidates for establishing content validity. Each of the business decision makers was
given five envelopes, each stating the name and definition of the substrata, as well as an
envelope for items that did not fit the substrata. Additionally, the participants were given
index cards with each prospective measurement item and asked to both sort and
categorize the items into the appropriate envelope. Based on the results of this exercise,
measurement items that had a majority agreement in categorization were retained and
ordered based on the averaged ranks. The items resulting from this exercise were
combined into a single instrument and a 7-point Likert scale (Likert, 1932) was added to
measure agreement with each item. See Table 13-16 for results of the categorization and
prioritization exercise.
Table 13. Content Validity Results of Geospatial Orientation and Navigation.
Initial Item Sort Results New Item Rank Rank Results
SPGS1 SPGON3
SPGS2 SPGON11 Removed
SPGS3 SPGON7
SPGS4 SPGON1
SPGS5 SPGON2
SPGS6 SPGON8
SPGS7 Removed
SPGS8 SPGON5
SPGS9 SPGON9
SPGS10 SPGON6
SPGS11 SPGON4
SPGS12 SPGON10
SPGS13 Removed
SPGS14 Removed
65
Table 14. Content Validity Results of Geospatial Memorization and Recall.
Initial Item Sort Results New Item Rank Rank Results
SPGMR1 SPGMR2
SPGMR2 SPGMR4
SPGMR3 SPGMR1
SPGMR4 SPGMR8
SPGMR5 SPGMR6
SPGMR6 SPGMR9
SPGMR7 SPGMR3
SPGMR8 SPGMR5
SPGMR9 SPGMR7
SPGMR10 SPGMR10
SPGMR11 SPGMR13 Removed
SPGMR12 SPGMR12 Removed
SPGMR13 SPGMR11 Removed
SPGMR14 Removed
66
Table 15. Content Validity Results of Geospatial Visualization.
Initial Item Sort Results New Item Rank Rank Results
SPGV1 SPGV4
SPGV2 SPGV7
SPGV3 SPGV2
SPGV4 SPGV6
SPGV5 SPGV1
SPGV6 SPGV13 Removed
SPGV7 SPGV12 Removed
SPGV8 SPGV8
SPGV9 SPGV5
SPGV10 SPGV3
SPGV11 SPGV14 Removed
SPGV12 SPGV11 Removed
SPGV13 SPGV10
SPGV14 SPGV9
Table 16. Content Validity Results of Geospatial Schematization.
Initial Item Sort Results New Item Rank Rank Results
SPGS1 SPGS7
SPGS2 SPGS11 Removed
SPGS3 SPGS1
SPGS4 Removed
SPGS5 SPGS6
SPGS6 SPGS10
SPGS7 SPGS9
SPGS8 SPGS8
SPGS9 SPGS4
SPGS10 SPGS12 Removed
SPGS11 SPGS2
SPGS12 Removed
SPGS13 SPGS5
SPGS14 SPGS3
67
Measurement Model Specification (Step 4)
The fourth step in the scale development procedure recommended by MacKenzie
et al. (2011) is to formally specify the measurement model. In this case, the relationship
between each proposed substratum and the GRA construct are visually and formally
presented.
Prior research has demonstrated that orientation and navigation are indicators of
spatial ability. For example, Kozlowski and Bryant (1977) revealed that perceptions of an
individual’s sense of direction reflected their spatial orientation ability. Also, Swink and
Speier (1999) discovered a relationship between spatial orientation and decision-
performance. Furthermore, Cherney et al. (2008) reported that spatial task performance
might be influenced by self-perceptions of navigation ability. As spatial orientation and
navigation appear to be indicators of spatial ability within a geospatial context, we
propose that:
Proposition P1: Self-perceived geospatial orientation and navigation ability
(SPGON) forms geospatial reasoning ability (GRA).
Additionally, research has demonstrated that memorization and recall are
indicators of spatial ability. In their study, Lei et al. (2009), discovered that subjects who
were familiar with specific landmarks were more likely to locate these landmarks using a
geospatial tool than with landmarks unfamiliar to them. Furthermore, Miyake et al.
(2001) discovered that the ability to memorize visiospatial concepts might influence
spatial ability. Additionally, subjects who had to retrace directions from a map reported
that during memorization they first translated the map into walking directions (Meilinger
& Knauff, 2008). Existing spatial tests include a visual memory test, which asks
68
participants to memorize an array of shapes on a page, and then identify these shapes
from recall (Velez et al., 2005). Thus, we propose:
Proposition P2: Self-perceived geospatial memorization and recall ability
(SPGMR) forms geospatial reasoning ability (GRA).
Furthermore, research has demonstrated that visualization ability is an indicator of
spatial ability. For example, Eisenberg and McGinty (1977) utilized a spatial
visualization test to measure spatial reasoning ability. Additionally, Velez et al. (2005)
discovered that spatial ability is correlated to the ability of 3-dimensional visualization.
Thus, we propose:
Proposition P3: Self-perceived geospatial visualization ability (SPGV) forms
geospatial reasoning ability (GRA).
Finally, researchers suggest that reducing the amount of information presented in
geovisualization to only include essential information can improve decision-making
performance. The benefits of such simplified maps is demonstrated by Agrawala and
Stolte (2001) who collected feedback from over 2,000 users of a technology that emulates
hand-drawn driving directions, which often emphasize essential information while
eliminating nonessential details. Additionally, Klippel et al. (2005) suggest that modern
cartographers can successfully develop schematic maps that are simplified, yet sufficient,
representations. In their study, Tversky and Lee (1998) asked students to provide both
written directions and a route map, of which 86% were able to provide sufficient written
directions and 100% were able to create sufficient route maps. Furthermore, Meilinger
and Knauff (2008) suggested that a reason subjects have difficulty regaining their
orientation, once lost, is that highly schematized information that does not contain
69
enough information for immediate orientation. No research was identified that
specifically recognized schematization ability as an indicator of reasoning ability,
however several studies indicated its importance for communicating geospatial
information (e.g., Klippel et al., 2005). As the relationship between SPGS and spatial
reasoning has not been established in literature, this construct was tested as a substratum
of GRA, as well as an independent construct. As the ability to effectively interpret
schematized geospatial information may influence GRA, we propose:
Proposition P4: Self-perceived geospatial schematization ability (SPGS) forms
geospatial reasoning ability (GRA).
As the CFT suggests that certain presentation modes facilitate decision-making,
based on an individual’s cognitive capability (Vessey, 1991; Speier, 2006), the substrata
presented above were essential to fully determine the effects of geovisualization on
decision-performance.
Reflective and Formative Constructs
While self-perception measurement items have commonly been used as reflective
measures of first-order constructs (e.g., Davis, 1989), there is a continuing debate
concerning the use of reflective or formative measurements for second-order
multidimensional constructs, such as GRA (e.g., Vlachos & Theotokis, 2009; Polites,
Roberts & Thatcher, 2012). Furthermore, Vlachos & Theotokis have discovered that
incorrectly specifying a second-order measurement model’s relationships can lead to
differing research conclusions.
In their criteria for distinguishing between reflective- and formative-indicator
models, MacKenzie, Podsakoff and Jarvis (2005) and MacKenzie et al. (2011) suggest
70
that constructs should be specified as reflective when (1) indicators are manifestations of
the construct, (2) changes in the indicator should not cause changes in the construct, (3)
changes in the construct do cause changes in the indicators, (4) dropping an indicator
should not alter the conceptual domain of the construct, (5) indicators are viewed as
affected by the same underlying construct and are parallel measures that co-vary, and (6)
indicators are required to have the same antecedents and consequences and to have a high
internal consistency and reliability. Based on this definition of a reflective construct, we
propose that the first-order constructs be modeled as reflective.
MacKenzie et al. (2011) further suggest that a construct should be specified as
formative when (1) the indicators are defining characteristics of the construct, (2)
changes in the indicators should cause changes in the construct, (3) changes in the
construct do not cause changes in the indicator, (4) dropping an indicator may alter the
conceptual domain of the construct, (5) it is not necessary for indicators to co-vary with
each other, and (6) indicators are not required to have the same antecedents and
consequences, nor have high internal consistency or reliability. Based on this definition
of a formative construct, we propose that the second-order construct be modeled as
formative.
Thus, subsequently the measurement model is referred to as a first-order
reflective, second-order formative model.
Visual Representation of Measurement Model
The measurement model is presented visually in Figure 3, demonstrating the ten
proposed measurement items for each first-order factor in addition to the propositions
that SPGON, SPGRM and SPGV are indicators of the GRA construct. Note that SPGS
71
has been included as a potential substratum of GRA in this model. While no previous
research has utilized schematization as an indicator of spatial ability, there is evidence
that SPGS could contribute to GRA. Thus, the authors felt that its inclusion was essential
to test its relationship to GRA.
SPGON
SPGV
SPGS
GRA
P1
P2
P4
SPGMR
P3
SPGON[item 1-10]
SPGMR[item 1-10]
SPGV[item 1-10]
SPGS[item 1-10]
Figure 3. First-Order Reflective, Second-Order Formative Measurement Model
including Propositions.
Mathematical Notation of Measurement Model
Each substratum is considered to be a latent (i.e., unobservable) variable,
measured using ten manifest (i.e., observable) variables. Thus, the relationship between
each measurement item and the substrata can be formally expressed as:
χ[1i] = λ[1i] X1 + ε[1i]
72
χ[2i] = λ[2i] X2 + ε[2i]
χ[3i] = λ[3i] X3 + ε[3i]
χ[4i] = λ[4i] X4 + ε[4i]
In the expressions above, χ[1i-4i] represent the measurement items for each
substratum, X[1-4] represent the four substrata, λ[1i-4i] represent the effect of X[1-4] on
χ[1i-4i], while ε[1i-4i] represent the measurement error for each indicator.
Furthermore, the GRA construct is a second-order formative construct defined by
the substrata. Thus, this relationship can be formally expressed as:
Y = Σ γ[1-4] χ[1-4] + ζ
In this expression, Y represents GRA, γ[1-4] represents the weigh associated with
each indicator substratum, χ[1-4] represents each indicator substratum and ζ represents
the common error term.
Data Collection (Step 5)
The next step of the scale development procedure recommended by MacKenzie et al.
(2011) was to perform the data collection. As this study performed two distinct pretests,
two data collection methods are presented.
Pretest A
Pretest A consisted of a small sample pretest, allowing the researchers to test
reliability and to gain initial feedback on the measurement items. For this pretest, 33
student subjects from an urban research university were asked to complete an online
survey and provide feedback.
73
Pretest B
In an attempt to reduce response bias, a variety of methods were used to recruit
subjects for Pretest B. These included contacting subjects in person, as well as through
social networks and e-mail. Since a variety of solicitation methods were used, there is no
meaningful way to show response rate.
For a traditional factor analysis, Hair, Black, Babin and Anderson (2010) suggest
a five-to-one ratio of the number of observations to number of variables. However, it is
further suggested that a more preferable sample would utilize a ten-to-one ratio. Knowing
this, and that our study contained 40 initial variables, our goal was to utilize a sample size
of at least 200, but preferably 400, subjects. Throughout a five-month period, 624
responses were collected. Of these, 24 incomplete observations were removed, retaining
600 responses. This number meets the minimum requirements for traditional factor
analysis and exceeds the minimum number necessary for a successful Partial Least
Squares analysis. Using Microsoft Excel 2010, random numbers were generated for each
row and the data was sorted based on this column. The sample was split and 300 samples
were retained for a validation study, while the other 300 samples were used for the
Pretest B analysis.
Scale Purification and Refinement (Step 6)
The next step recommended by MacKenzie et al. (2011) is to purify and refine the
scale. This step was accomplished by performing two pretests.
Pretest A
For Pretest A, a small sample pretest of the research instrument was conducted. For
this pretest, 33 subjects were asked to complete an online survey, measuring the results of
74
all 40 substrata measurement items. Additionally, the subjects were given an opportunity
to provide input via open-ended questions designed to elicit feedback on the actual
measurement items. The open-ended question responses were reviewed for additional
opportunities of scale refinement. An Item Analysis performed using IBM SPSS
Statistics 19 revealed either good or acceptable levels of Cronbach’s alpha (as shown in
Table 17) for each substratum and SPGS, (Cronbach, 1951; Gliem & Gliem, 2003).
Furthermore, the open-ended feedback demonstrated that the questions were clear and
interesting to the participants. This was crucial as one of the goals identified for the
measurement scale was that it could successfully determine GRA in subjects who do not
necessarily have a familiarity with geographic concepts.
Table 17. Pretest A Reliability Statistics.
Substrata Cronbach’s Alpha No of Items No of Subjects
SPGON .830 10 33
SPGMR .846 10 33
SPGV .835 10 33
SPGS .735 10 33
Pretest B
Based on the success of Pretest A, an exploratory data analysis using Partial Least
Squares (PLS) Structural Equation Modeling (SEM) was performed.
Given that the selected analytical methodology can influence the results, great
care was taken to ensure the most appropriate statistical analysis method. Polites et al.
(2012) determined that between the years 2000 and 2009, 84% of studies testing
aggregate constructs utilized PLS, indicating that the use of PLS in IS research has
become widely accepted. Furthermore, Ringle, Götz, Wetzels and Wilson (2009)
75
suggested the use of PLS path modeling over maximum likelihood covariance-based
structural equation modeling for tests of second-order constructs when methodological
requirements are not met. As the primary goal of this study was to define a new construct
on an untested measurement model, PLS path modeling was utilized to ensure consistent
results. As such, SmartPLS (Ringle, Wende, & Will, 2005) was utilized for subsequent
measurement model testing.
The results of this analysis helped establish construct validity and provided an
opportunity for refinement of the measurement scale. The first goal of Pretest B was to
identify parsimonious sets of variables. Furthermore, a second goal was to reduce the
number of items significantly and to identify only the most representative items for each
construct. The pretest procedure resulted in a reduction in measurement items from 40 to
12. Additionally, the pretest revealed that SPGS was not a valid indicator of GRA.
The Pretest B sample population (n=300) included demographic variables
measuring age, gender, education, profession and cultural background. The results of
these demographic items are presented in Table 18.
76
Table 18. Pretest B Descriptive Statistics of Demographics Variables.
Question Variables Percentage
Age 18-25 30.7
26-35 32.3
36-45 12.0
45-55 12.3
56-65 9.0
66 3.7
Gender Female 60.0
Male 40.0
Education 2 Year/Associate Degree 24.7
4 Year/Bachelor Degree 30.7
Doctor/JD/PhD 2.7
Elementary/Middle School 0.7
High School 30.7
Master Degree 10.7
Profession Business Professional 43.7
Student 37.7
Geospatial Professional 0.7
Other 18.0
Culture African 3.7
Asian 8.0
Australian 0.3
European 38.7
Middle Eastern 4.3
North American 43.7
South American 1.3
The first analysis of Pretest B involved a PLS factor analysis using SmartPLS
(Ringle et al., 2005). For this analysis the measurement items of each first-order
substratum were modeled as reflective. The first step was to test the full research model,
which included SPGS as a substratum of GRA. This model revealed that the path
77
coefficient between SPGS and GRA was 0.047, while the remaining path coefficients
ranged from 0.355 to 0.376 (see Figure 4). This agreed with prior research, which had not
included schematization as an indicator of spatial reasoning. However, as there is
evidence that SPGS could contribute to GRA in some way (Agrawala & Stolte, 2001;
Klippel et al., 2005), it was further refined using an IBM SPSS Statistics 19 factor
analysis. Another reason to have continued the development of the SPGS construct is that
it is a construct similar to the GRA measure, so it can be used to establish discriminate
validity in future testing.
SPGON
SPGMR
SPGS
GRA
0.376
0.394
0.04
7
SPGMR
0.355
Figure 4. Pretest B, Path Analysis including SPGS.
As prior research has shown that an equal number of indicator variables are
preferred for first-order factors using multiple indicators, only the four highest loading
items were retained. Figure 5 presents the path analysis findings and Table 19 presents
cross-loadings for each of the substrata.
78
SPGON6
SPGON
SPGMR
SPGV
SPGON7 SPGON4
SPGON5
SPGMR3
SPGMR2
SPGMR10
SPGMR9
GRA
0.8104
0.382
0.299
0.55
80
0.5193
0.7485
0.70
24
0.5098
0.6160
0.8045
0.6179
0.69
590.
2978
0.5
86
1 0.6
98
0
SPGMR4
SPGMR6
SPGMR7
SPGMR8
SPGMR1
SPGMR5
SPGON8 SPGON3
SPGON9 SPGON2
SPGON10 SPGON1
SPGV6 SPGV5
SPGV7 SPGV4
SPGV8 SPGV3
SPGV9 SPGV2
SPGV10 SPGV1
0.396
0.8043
0.8233
0.8473
0.7597
0.5875
0.7760
0.7306
0.6
97
00.7939
0.77720.7910
0.7
44
9 0.8
39
20.
8682
0.82
120.
7882
0.69
38
Figure 5. Pretest B, Path Analysis including SPGS.
79
Table 19. PLS Factor Analysis Loadings.
Factor Loading by Substratum and GRA
Item SPGMR SPGON SPGV GRA
SPGMR1 0.8045 0.6625 0.5896 0.748
SPGMR2 0.8043 0.6242 0.556 0.7233
SPGMR3 0.8233 0.7056 0.6486 0.7896
SPGMR4 0.8473 0.7 0.6772 0.8055
SPGMR5 0.7597 0.6439 0.5544 0.7126
SPGMR6 0.5875 0.4979 0.4109 0.5457
SPGMR7 0.776 0.6584 0.7327 0.778
SPGMR8 0.5193 0.4697 0.3074 0.477
SPGMR9 0.7306 0.6572 0.5596 0.7078
SPGMR10 0.7485 0.6817 0.6481 0.7507
SPGON1 0.551 0.6938 0.5459 0.6465
SPGON2 0.6474 0.7882 0.5283 0.7155
SPGON3 0.6569 0.8212 0.6022 0.754
SPGON4 0.757 0.8682 0.601 0.8112
SPGON5 0.7406 0.8392 0.6156 0.798
SPGON6 0.6611 0.7449 0.5025 0.6967
SPGON7 0.4304 0.5098 0.3377 0.4662
SPGON8 0.4832 0.616 0.5218 0.5827
SPGON9 0.539 0.6179 0.5253 0.6066
SPGON10 0.7298 0.8104 0.5752 0.7707
SPGV1 0.6616 0.6594 0.791 0.7505
SPGV2 0.7018 0.6204 0.7772 0.7474
SPGV3 0.675 0.6072 0.7939 0.7367
SPGV4 0.5709 0.5281 0.697 0.6363
SPGV5 0.5091 0.462 0.698 0.5869
SPGV6 0.3526 0.337 0.5861 0.4437
SPGV7 0.2337 0.2504 0.2978 0.2773
SPGV8 0.4771 0.4669 0.6959 0.5754
SPGV9 0.3752 0.3499 0.558 0.4492
SPGV10 0.4642 0.4675 0.7024 0.5725
Highest Item Loadings shown in bold.
80
The latent variable correlation, as shown in Table 20, shows the correlations
between GRA and its substrata. Each substratum correlates with GRA between 0.8852
and 0.9528, while there is also a high correlation between each of the substrata.
Table 20. Latent Variable Correlation.
SPGMR SPGON SPGV GRA
SPGMR 1 - - -
SPGON 0.8499 1 - -
SPGV 0.776 0.7298 1 -
GRA 0.9528 0.9369 0.8852 1
As demonstrated in Table 21, the path coefficients are relatively equal and range
between 0.299 and 0.396. Relatively equal path coefficients are a suggested requirement
of research models using repeating indicators (Beemer & Gregg, 2010). Additionally, all
first-order constructs achieve t-statistic values greater than 1.96 demonstrating
convergent validity (Gefen & Straub, 2005). Furthermore, a comparison of the first-order
latent variable correlations reveals that all are below 0.9, indicating no common method
bias (Bagozzi, Yi, & Phillips, 1991).
Table 21. Path Coefficients.
Original
Sample
Sample
Mean
Standard
Deviation
Standard
Error
T
Statistics
SPGMR -> GRA 0.396 0.395 0.0129 0.0129 30.7691
SPGON -> GRA 0.382 0.3813 0.0112 0.0112 34.1738
SPGV -> GRA 0.299 0.2995 0.0143 0.0143 20.9492
Next, construct reliability and validity were established. The Average Variance
Extracted (AVE) for a first-order construct should be greater than 0.50 (Fornell &
Larcker, 1981), which SPGRM and SPGON achieve, and thus support, construct validity.
However, SPGV only reached an AVE of 0.4554, which is a concern. Once non-essential
81
or incorrectly loading items are removed this value may improve and will need to be re-
evaluated. For second-order constructs with formative indicators, such as GRA, a R2
value greater than 0.5 can indicate construct validity (Diamantopoulos, Reifler, & Roth,
2008; MacKenzie et al., 2011). In the measurement model, the path analysis revealed a
R2 value of 1.0 as all indicator variables also defined the second-order construct. The
composite reliability scores were greater than 0.8888 for each of the substrata, which
meant that reliability as defined by Cronbach (1951) and Fornell and Larcker for
reflective, first-order constructs, was achieved. Gleim and Gleim (2003) suggest a
minimum alpha of 0.8 as a reasonable goal, which each of the tested substrata exceeded.
Per MacKenzie et al. as the measurement model does not predict correlation of the sub-
dimensions for the second-order, formative construct of GRA, reliability is not relevant
for this construct. See Table 22 for more detail.
Table 22. Construct Reliability.
AVE Composite
Reliability
R
Square
Cronbach’s
Alpha
Communality Redundancy
GRA 0.4468 0.9588 1 0.9546 0.4468 0.2671
SPGRM 0.5578 0.9253 0 0.909 0.5578 0
SPGON 0.5468 0.9218 0 0.9037 0.5468 0
SPGV 0.4554 0.8888 0 0.86 0.4554 0
Next, items that did not load highly upon their corresponding substrata were
removed. Indicated by prior research, an equal number of indicators should be utilized for
all first order constructs, therefore only the four highest loading measurement items were
retained. Care was taken to ensure that an equal number of measurement items remained
for each substratum, as this would ensure the integrity of the model. The item reduction
82
for SPGS occurred using a traditional factor analysis using IBM SPSS Statistics 19 and
five items were retained.
Pretest B – Reduced Items
Once the items were reduced, an additional path analysis of Pretest B was
performed using SmartPLS (Ringle et al., 2005). Again, the measurement items of each
first-order substratum were modeled as reflective, while the second-order construct of
GRA was modeled using the hierarchical component model. This model, along with the
updated path coefficients, is presented in Figure 6. The reduced items for each substratum
are presented in Table 23 along with factor loadings.
SPGON3
SPGON
SPGMR
SPGV
SPGON4
SPGON5
SPGON10
SPGMR1
SPGMR2
SPGMR3
SPGMR4
SPGV1
SPGV2
SPGV3
SPGV10
GRA
0.3901
0.3798
0.33
82
0.8439
0.9201
0.8772
0.8539
0.8492
0.8760
0.8443
0.8766
0.8509
0.9018
0.9076
0.6902
Figure 6. Pretest B, Path Analysis.
83
Table 23. PLS Factor Analysis Loadings.
Factor Loading by Substratum and GRA
Item SPGMR SPGON SPGV GRA
SPGMR1 0.8492 0.6507 0.5601 0.7658
SPGMR2 0.876 0.6396 0.5372 0.7639
SPGMR3 0.8443 0.6756 0.6837 0.8155
SPGMR4 0.8766 0.6685 0.6396 0.81
SPGON3 0.6287 0.8439 0.5732 0.7619
SPGON4 0.6893 0.9201 0.6166 0.8293
SPGON5 0.6722 0.8772 0.6332 0.8116
SPGON10 0.6835 0.8539 0.5812 0.7892
SPGV1 0.6167 0.63 0.8509 0.7678
SPGV2 0.6579 0.6128 0.9018 0.7939
SPGV3 0.6291 0.6059 0.9076 0.7822
SPGV10 0.4078 0.4147 0.6902 0.5332
The latent variable correlation, as shown in Table 24, shows that the correlations
between GRA and each of its substrata after the items for each substratum were reduced.
Each substratum correlates with GRA between 0.8742 and 0.9165, while there is also a
high correlation between each of the substrata.
Table 24. PLS Latent Variable Correlation.
SPGMR SPGON SPGV GRA
SPGMR 1 - - -
SPGON 0.765 1 - -
SPGV 0.7045 0.688 1 -
GRA 0.9165 0.9133 0.8742 1
After reducing the number of items from ten to four for each substratum, the path
coefficients, as shown in Table 25, continue to be relatively equal and range between
0.3382 and 0.3901. As such relative similarity was a suggested requirement of research
models using repeating indicators (Beemer & Gregg, 2010) we deem the results
84
acceptable. Additionally, convergent validity can be demonstrated when t-statistic values
are greater than 1.96 (Gefen & Straub, 2005), which all of the first-order constructs
achieve. To ensure that there was no systematic influence biasing the data, two common
method bias tests were performed. The first involved the addition of a theoretically
dissimilar marker construct to which each of the substrata was compared. Using this
method, the greatest squared variance revealed only a 2.87% shared variance, below the
3% suggested threshold (Lindell & Whitney, 2001). Furthermore, a comparison of the
first-order latent variable correlations reveals that all are below 0.9, indicating no
common method bias (Bagozzi et al., 1991).
Table 25. Path Coefficients.
Original
Sample
Sample
Mean
Standard
Deviation
Standard
Error
T Statistics
SPGMR -> GRA 0.3798 0.3795 0.0114 0.0114 33.3878
SPGON -> GRA 0.3901 0.3899 0.0096 0.0096 40.6855
SPGV -> GRA 0.3382 0.3387 0.011 0.011 30.8816
Next, construct reliability and validity, were further established. The Average
Variance Extracted (AVE) for a first-order construct should be greater than 0.5 (Fornell
& Larcker, 1981), which all substrata achieve. The concern encountered when analyzing
the AVE of the substrata containing all measurement items in Pretest B was resolved
through the item reduction. For second-order constructs with formative indicators, such
as GRA, a R2 value greater than 0.5 indicates construct validity (Diamantopoulos et al.,
2008; MacKenzie et al., 2011). In the measurement model the path analysis revealed a R2
value of 1.0 as all indicator variables defined the second-order construct. The composite
reliability scores were greater than 0.8986 for each of the substrata, which meant that
reliability as defined by Cronbach (1951) and Fornell and Larcker (1981) for reflective,
85
first-order constructs, was achieved. Gleim and Gleim (2003) suggest a minimum alpha
of 0.8 as a reasonable goal, which each of the tested substrata exceeded. Per MacKenzie
(2011), as the measurement model does not predict correlation of the sub-dimensions for
the second-order, formative construct of GRA, reliability is not relevant for this
construct. See Table 26 for more detail.
Table 26. Construct Reliability.
AVE Composite
Reliability
R
Square
Cronbach’s
Alpha
Communality Redundancy
GRA 0.5964 0.9461 1 0.9372 0.5964 0.3291
SPGRM 0.7425 0.9202 0 0.8843 0.7425 0
SPGON 0.7643 0.9283 0 0.8968 0.7643 0
SPGV 0.6927 0.8986 0 0.8465 0.6927 0
Finally, discriminant and convergent validity were established. Convergent
validity is demonstrated as all items that should load in GRA did so at greater than 0.6,
except for SPGV10 that loaded slightly below 0.6 at 0.5779. Particular attention should
be given to this item in future uses of this measurement scale to ensure continued
validity. Additionally, all items that should load on SPGS, which is now treated as an
independent construct, did so greater than 0.6. Furthermore, discriminate validity was
further established as GRA substrata items that should not load on SPGS based on theory,
only do so at less than 0.3. See Table 27 for more detail.
86
Table 27. PLS Factor Loading and Cross Loading.
GRA SPGS
GRA1 (SPGMR1) 0.8038 0.2268
GRA2 (SPGMR2) 0.7827 0.1875
GRA3 (SPGMR3) 0.7878 0.0983
GRA4 (SPGMR4) 0.8091 0.1407
GRA5 (SPGON3) 0.7864 0.1957
GRA6 (SPGON4) 0.8204 0.1299
GRA7 (SPGON5) 0.8004 0.1253
GRA8 (SPGON10) 0.7948 0.1656
GRA9 (SPGV1) 0.7454 0.1167
GRA10 (SPGV2) 0.7351 0.0574
GRA11 (SPGV3) 0.7263 0.0829
GRA12 (SPGV10) 0.5779 0.1938
SPGS4 0.2026 0.8631
SPGS5 0.1982 0.7975
SPGS8 0.1578 0.7485
SPGS9 0.1258 0.7748
SPGS10 0.0848 0.7398
Using the two-step procedure outlined by Gefen and Straub (2005), discriminant
validity was further established by comparing inter-construct correlations with square
root of each construct’s AVE. Gefen and Straub state that the square root of the AVE
should be ‘much larger’ than the construct correlations. Chin (1998) suggests that
discriminant validity can be inferred when the variance of each construct is larger than
the variance shared with any other construct. Furthermore, Fornell and Larcker (1981)
suggest that all AVEs should exceed a 0.50 threshold, which occurs in this analysis.
Based on the results of this analysis, as seen in Table 28, the square root of the AVE for
each substratum is larger than any correlation.
87
Table 28. Inter-Construct Correlations and Square Root of AVE.
SPGON SPGMR SPGV Square Root of AVE
SPGON - 0.765 0.688 0.8742
SPGMR 0.765 - 0.7045 0.8617
SPGV 0.688 0.7045 - 0.8323
The analysis performed allowed for construct reliability and validity, discriminate
reliability, indicator reliability and convergent reliability to be established. Next, the scale
will be reassessed using a second data set.
Data Collection and Reexamination of Scale Properties (Step 7)
The seventh step, as recommended by MacKenzie et al. (2011), is to collect
additional data and to use this data to reexamine the scale derived in the previous steps.
The three-hundred responses from the initial data collection effort were utilized for this
step. The demographic information on the second data set is provided in Table 29.
88
Table 29. Study 1 Descriptive Statistics of Demographic Variables.
Question Variables Percentage
Age 18-25 32.7
26-35 32.0
36-45 12.0
45-55 12.7
56-65 6.7
66 4.0
Gender Female 52.0
Male 48.0
Education 2 Year/Associate Degree 26.3
4 Year/Bachelor Degree 39.3
Doctor/JD/PhD 3.0
Elementary/Middle School 1.0
High School 30.3
Master Degree 10.0
Profession Business Professional 51.7
Student 33.3
Geospatial Professional .7
Other 14.3
Culture African 3.7
Asian 13.7
Australian 0
European 33.0
Middle Eastern 4.0
North American 44.7
South American 1.0
As with Pretest 2, the Test data was analyzed using SmartPLS (Ringle et al.,
2005). The path coefficients of the first-order constructs (substrata) on the second-order
construct appeared to be relatively equal, an important consideration for a multi-
dimensional second-order construct. Again, the measurement items of each first-order
89
substratum were modeled as reflective, while the second-order construct of GRA was
modeled using the hierarchical component model. This model is presented in Figure 7
along with the Test’s path coefficients. The items and their factor loadings are presented
in Table 30.
SPGON3
SPGON
SPGMR
SPGV
SPGON4
SPGON5
SPGON10
SPGMR1
SPGMR2
SPGMR3
SPGMR4
SPGV1
SPGV2
SPGV3
SPGV10
GRA
0.3874
0.3938
0.34
37
0.8684
0.9361
0.8942
0.8099
0.8746
0.9186
0.8666
0.8498
0.8575
0.9077
0.8977
0.7018
Figure 7. Test, Path Analysis.
90
Table 30. PLS Factor Analysis Loadings and Weights.
Factor Loading by Substratum and GRA
Item SPGMR SPGON SPGV GRA
SPGMR1 0.8746 0.6908 0.5779 0.8106
SPGMR2 0.9186 0.6642 0.5629 0.8125
SPGMR3 0.8666 0.609 0.6846 0.8125
SPGMR4 0.8498 0.598 0.5576 0.7579
SPGON3 0.58 0.8684 0.5419 0.751
SPGON4 0.6798 0.9361 0.6128 0.8409
SPGON5 0.6621 0.8942 0.5275 0.7884
SPGON10 0.6405 0.8099 0.5577 0.7576
SPGV1 0.6551 0.6011 0.8575 0.7856
SPGV2 0.5753 0.5513 0.9077 0.7521
SPGV3 0.5554 0.5174 0.8977 0.7277
SPGV10 0.499 0.4794 0.7018 0.6234
The latent variable correlation, as shown in Table 31, demonstrates that each
substratum correlates with GRA between 0.8586 and 0.9102, while there is also a high
correlation (greater than 0.63) between each of the substrata.
Table 31. Latent Variable Correlation.
SPGMR SPGON SPGV GRA
SPGMR 1 - - -
SPGON 0.7304 1 - -
SPGV 0.6794 0.6384 1 -
GRA 0.9102 0.8944 0.8586 1
The path coefficients, as shown in Table 32, continue to be relatively equal and
range between 0.3437 and 0.3948 (while they ranged from 0.3382 and 0.3901 in pretest
B). Relatively similar path coefficients were a suggested requirement of second-order
formative research models using repeating indicators. Additionally, convergent validity is
demonstrated when t-statistic values are greater than 1.96 (Gefen and Straub, 2005),
which was exceeded by all first-order constructs. To ensure that there was no systematic
91
influence biasing the data, two common method bias tests were performed. The first
involved the addition of a theoretically dissimilar marker construct to which each of the
substrata was compared. Using this method, the greatest squared variance revealed only a
2.87% shared variance, below the 3% suggested threshold (Lindell & Whitney, 2001).
Furthermore, a comparison of the first-order latent variable correlations reveals that all
are below 0.9, indicating no common method bias (Bagozzi et al., 1991).
Table 32. Path Coefficients.
Original
Sample
Sample
Mean
Standard
Deviation
Standard
Error
T
Statistics
SPGMR -> GRA 0.3938 0.3948 0.013 0.013 30.2249
SPGON -> GRA 0.3874 0.3865 0.0131 0.0131 29.5689
SPGV -> GRA 0.3437 0.3438 0.0128 0.0128 26.8764
Next, construct reliability and validity, as well as convergent validity, were
established. The AVE for a first-order construct should be greater than the 0.5 threshold
(Fornell & Larcker, 1981), which all three substrata exceed at values greater than 0.7. For
second-order constructs with formative indicators, such as GRA, a R2 value greater than
0.5 indicates construct validity (Diamantopoulos et al., 2008; MacKenzie et al., 2011). In
the measurement model the path analysis revealed a R2 value of 1.0 as all indicator
variables defined the second-order construct. The composite reliability scores were
greater than 0.9 for each of the substrata, which was an improvement over the pretest
results. See Table 33 for the full results.
92
Table 33. Construct Reliability.
AVE Composite
Reliability
R
Square
Cronbach’s
Alpha
Communality Redundancy
GRA 0.5933 0.9457 1 0.937 0.5933 0.3333
SPGRM 0.7705 0.9306 0 0.9004 0.7705 0
SPGON 0.7715 0.9309 0 0.9001 0.7715 0
SPGV 0.7144 0.9083 0 0.8629 0.7144 0
Finally, discriminant and convergent validity were established. All items that
should load on GRA did so at greater than 0.6, which helped establish convergent
validity. SPGV10, which was a potential concern in the previous test, improved to a more
acceptable 0.632. Additionally, all items that should load on SPGS did so at greater than
0.6. Furthermore, as the SPGS items that should not load on GRA, based on a lack of
such findings in existing literature, did so at less than 0.2, which was also an
improvement over the pretest results, which contributes to discriminate validity. See
Table 34.
93
Table 34. PLS Factor Loading and Cross Loading.
GRA SPGS
GRA1 (SPGMR1) 0.8197 0.1321
GRA2 (SPGMR2) 0.8305 0.1064
GRA3 (SPGMR3) 0.8274 0.1778
GRA4 (SPGMR4) 0.7927 0.1762
GRA5 (SPGON3) 0.7539 0.1429
GRA6 (SPGON4) 0.8276 0.0747
GRA7 (SPGON5) 0.7903 0.1309
GRA8 (SPGON10) 0.7461 0.0967
GRA9 (SPGV1) 0.7726 0.12
GRA10 (SPGV2) 0.7071 0.0701
GRA11 (SPGV3) 0.677 0.0401
GRA12 (SPGV10) 0.632 0.1181
SPGS4 0.1462 0.7953
SPGS5 0.0597 0.6452
SPGS8 0.1482 0.745
SPGS9 0.097 0.8084
SPGS10 0.1355 0.7702
Using the two-step procedure outlined by Gefen and Straub (2005), discriminant
validity was further established by comparing inter-construct correlations with the square
root of each construct’s AVE. Gefen and Straub (2005) state that the square root of the
AVE should be ‘much larger’ than the construct correlations. Chin (1998) suggests that
discriminant validity can be inferred when the variance of each construct is larger than
the variance shared with any other construct. Furthermore, Fornell and Larcker (1981)
suggest that all AVEs should exceed a 0.50 threshold, which occurs in this analysis.
Based on the results of this analysis, as seen in Table 35, the square root of the AVE for
each substratum is larger than any correlation. However, this preliminary analysis should
be confirmed in a future study by comparing correlations between a conceptually similar
construct and ideally discovering a correlation of less than 0.71, as was suggested by
MacKenzie et al. (2011).
94
Table 35. Inter-Construct Correlations and Square Root of AVE.
SPGON SPGMR SPGV Square Root of
AVE
SPGON - 0.7304 0.6384 0.8784
SPGMR 0.7304 - 0.6794 0.8778
SPGV 0.6384 0.6794 - 0.8452
Collinearity
Finally, due to the formative nature of the construct, an assessment of collinearity
was performed. Specifically, each of the substrata measurement items were examined for
tolerance and variance inflation factor (VIF), the reciprocal of tolerance. This test was
performed using IBM SPSS Statistics 19 and revealed that the VIF does not cross the
value threshold of 5.0 (Hair, Ringle & Sarstedt, 2011; Hair, Hult, Ringle & Sarstedt,
2014), which would indicate a potential collinearity problem.
Gender Concerns
Finally, as gender has been shown to cause differences in spatial cognitive
abilities (Albert & Golledge, 1999) it is important to determine if gender differences
influence the outcome of the test sample, thus additional analyses were performed using
subgroups consisting of only female (n=166) or male (n=134) subjects.
The first test addressed the effects of SPGON on GRA based on gender. This test
revealed regressions weights of 0.435 and 0.326 for female and male subjects,
respectively. To determine if this change was statistically significant a t-statistic and p-
value were calculated, revealing that the differences between female and male subjects
were indeed statistically significant at a p-value of 0.000. See Table 36 for more detail.
95
Table 36. SPGON GRA: Gender Group Differences.
Female Subjects Male Subjects
Sample Size 166 134
Regression Weight 0.435 0.326
Standard Error 0.019 0.0220
t-statistic 3.757
p-value (2-tailed) 0.000
The second test addressed the effects of SPGMR on GRA based on gender. This
test revealed regressions weights of 0.410 and 0.398 for female and male subjects,
respectively. While these regressing weights are very similar, a t-statistic and p-value
confirmed that there is no statistical significance between SPGMR and GRA based on
gender. See Table 37 for more detail.
Table 37. SPGMR GRA: Gender Group Differences.
Female Subjects Male Subjects
Sample Size 166 134
Regression Weight 0.410 0.398
Standard Error 0.0167 0.0178
t-statistic 0.491
p-value (2-tailed) 0.624
Finally, the effects of SPGV on GRA based on gender were analyzed. This test
revealed regressions weights of 0.310 and 0.399 for female and male subjects,
respectively. A t-statistic and p-value were calculated revealing that the differences
between female and male subjects were statistically significant at a p-value of 0.005. See
Table 38 for more detail.
96
Table 38. SPGV GRA: Gender Group Differences.
Female Subjects Male Subjects
Sample Size 166 134
Regression Weight 0.310 0.399
Standard Error 0.0194 0.0254
t-statistic 2.843
p-value (2-tailed) 0.005
These findings demonstrate that the SPGON and SPGV substrata encounter
statistically significant differences based on gender, which aligns with results indicating
that gender differences influenced the outcomes of spatial reasoning tests from previous
studies. This finding further indicates that the gender differences present in spatial
reasoning also find their way into geospatial reasoning. Of additional interest is that
SPGMR was not affected by gender.
Discussion
Based on the above analysis it appears that the propositions indicating that
SPGON, SPGMR and SPGV are substrata of GRA are in agreement with the proposed
measurement model. However, it was also revealed that SPGS is not an indicator of
GRA, but instead an independent construct. This follows existing literature as the
SPGON, SPGMR and SPGV substrata focus on an individuals’ self-perceived ability to
navigate space using existing environmental knowledge or tools such as maps, while the
SPGS construct essentially focuses on the communication of such knowledge. While it
could be expected that the average business decision-maker would utilize the concepts
included in SPGON, SPGMR and SPGV regularly, the SPGS concepts could be more
applicable to a geospatial professional, such as a cartographer. Such an individual would
97
use a process consisting of abstraction, idealization, and selection to develop schematized
geospatial information (Klippel et al., 2005; Herskovits, 1998) more often than a business
decision-maker. Further research will be essential to better understand the importance of
SPGS, as defined in this study, on business decision-making using geospatial data and
how it can be best applied to communicate information to decision-makers.
While we made some progress to better understand and define SPGS, to further
demonstrate the need for a comprehensive definition of schematization, Klippel et al.
(2005) suggest additional research and ask the following questions:
Is schematization a process or the result of a process? Are concepts or
conceptualizations the result of a schematization process? For the map domain
this may be easier as we could claim that the result of a schematization process is
a schematic map, which is a non-deniable fact, but what is the schematic map
composed of? Another question that should be answered is: do we schematize
spatial relations or do we schematize spatial objects? Do we schematize
intersections or do we schematize angles between streets? (p. 59)
Furthermore, unlike SPGON, SPGMR and SPGV whose underlying concepts
related to spatial ability have been successfully applied in numerous previous studies,
SPGS will require further research particularly to determine how much, if any, impact
there is on decision-performance.
Additionally, it was shown that gender causes statistically significant differences
in the measurement model, which was the case in previous literature. Table 39 presents
the final 12 items of the GRA measurement instrument utilizing the three substrata.
98
Table 39. Final GRA Measurement Items.
Item ID Item ID
SPGMR1 I can usually remember a new route after I have traveled it only once.
SPGMR2 I am good at giving driving directions from memory.
SPGMR3 After studying a map, I can often follow the route without needing to
look back at the map.
SPGMR4 I am good at giving walking directions from memory.
SPGON3 In most circumstances, I feel that I could quickly determine where I am
based on my surroundings.
SPGON4 I have a great sense of direction.
SPGON5 I feel that I can easily orientate myself in a new place.
SPGON10 I rarely get lost.
SPGV1 I can visualize geographic locations.
SPGV2 I can visualize a place from information that is provided by a map
without having been there.
SPGV3 I can visualize a place from a map.
SPGV10 While reading written walking directions, I often form a mental image
of the walk.
The final measurement items for SPGS are presented in Table 40.
Table 40. Final SPGS Measurement Items.
Item ID Item ID
SPGS4 I prefer maps that show only the most essential information.
SPGS5 I like simple, clear maps.
SPGS8 I prefer written walking directions that only include the most essential
navigational elements.
SPGS9 I prefer maps that only provide information necessary to accomplish
my tasks.
SPGS10 I prefer maps that only provide key information, even if they are not
to scale.
Expected implications to industry and research, as well as expected limitations of
the research are presented next.
99
Implication to Research
As prior research has revealed conflicting results on the importance of spatial
ability on decision performance, particularly when utilizing geographic data, the results
of this research will more narrowly define geospatial reasoning ability for business
decision-makers and provided an easy to administer instrument which can provide
researchers with a stronger, context-specific measure to be extended into future research.
Ideally, the GRA construct will be compared to existing measures, such as VZ-2, during
a decision-making task experiment. The measurement developed during this study will
allow future researchers to more easily incorporate geospatial reasoning ability into their
experiments, as laboratory settings and training evaluators are not necessary for the
evaluation of research results.
Implication to Industry
In addition to providing researchers with a valid construct and measurement scale,
this research could provide business leaders with a measurement tool to determine current
and future employee’s geospatial reasoning ability within a business context. The
researchers feel that this ability may be an essential component for the successful
interpretation of geovisualized data and could allow organizations to make better
decisions from such data. When building teams of decision-makers who will utilize
geospatial data to make their decisions, such as market researchers or site location
specialists, organizations could rely on future versions of the GRA scale to qualify
candidates. Furthermore, some industries, such as aviation and sea transport, rely on the
expectation that candidates possess strong geospatial reasoning skills. Using a measure,
100
such as GRA, these industries may be able to pre-qualify candidates before administering
additional laboratory testing.
Limitations and Future Research
While the current study represents only initial steps to establish a refined
measurement instrument, future studies will need to be conducted in order to further
establish the validity of GRA and its substrata. Future steps should explore the scale
validity further by performing experimental manipulation of the construct, establishing
nomological validity, establishing criterion validity, performing a known-groups
comparison, and further assessing the discriminant validity. Each of these tests is
essential to strengthening the validity of the GRA construct and its measurement scale.
As the construct proposed in this research project will most likely first be tested in
highly controlled experiments to ensure strong internal validity, it will be essential that
the construct be extended to ensure greater generalizability and provide external validity.
The topic of generalizability, or the ability to generalize research from one setting
to another, has been debated heavily in recent IS journals (e.g., Compeau et al., 2012; Lee
& Hubona, 2009; Lee & Baskerville, 2003, 2012; Tsang & Williams, 2012). Lee and
Baskerville (2003) suggest that a theory that lacks generalizability also lacks usefulness,
which is essential in a field that also has professional implications, such as IS. Tsang and
Williams disagreed with many of the assumptions in Lee and Baskerville’s (2003) work
and suggested defining the intended meaning of generalizability, as well as indicating to
what population a study is being generalized. In 2012, Lee and Baskerville recommend
that researchers perform four “judgment calls” when generalizing, including (1) a
judgment of the “uniformity of nature”, (2) a judgment of conditions between the study
101
setting and the generalization, (3) a judgment of successfully having included all
necessary variables to generalize and (4) assessing whether a theory is indeed true, which
would be a requirement to generalize a theory. Compeau et al. present, in a humorous
rhetorical device, that many authors address generalizability by including a paragraph
similar to the following during the third round of revisions:
The findings of this study must be considered in light of its limitations.
The use of students, rather than working professionals, is one potential limitation
of the results. It may be that the results are unique to a student population and
would not generalize to a population of working professionals. Nonetheless, we
are testing a broad theory of behavior so we see no reason to expect a difference.
Moreover, our goal in this paper was to develop theory and measures, rather than
to provide a definitive test of the theory, and thus we felt it was important to focus
on internal validity rather than external validity. (p. 1094)
Furthermore, Compeau et al. (2012) provide four recommendations for
generalizability, including (1) explicitly presenting the goal of the research, (2) explicitly
defining the population to which the study should generalize, (3) justifying the sample
choice, (4) discussing limitations of the sample choice and (5) providing consistency
between each of these elements. Based on these recommendations this chapter research
goal was to provide a general measure of GRA that applies to a global population of
information technology users. The sample choice of Internet users applies to this
population group; however, a clear limitation is that the sample may not include
sufficient representations of experience levels, cultures, etc. Thus, the results of this study
102
are expected to generalize to a larger group of technology users; however, this cannot be
stated as a fact.
To limit the potential of common method bias impacting construct validity,
several procedural techniques were utilized in the scale development. For example, as
suggested by Podsakoff, MacKenzie, Lee, & Podsakoff (2003), subject anonymity was
ensured and respondents were informed that there were no right or wrong answers, and
measurement items were sources from varied literature and sorted by both industry and
academic experts. Furthermore, the recommendations of Tourangeau, Rips, and Rasinski
(2000) were used, including, but not limited to, labeling all scale values, using simple and
concise questions, avoiding unfamiliar terms, and avoiding complicated syntax. While
such procedural remedies were utilized to minimize common method bias during the
scale development, another limitation of this study is the limited ability to test for
common method bias statistically. While various techniques exist, such as Harman’s
Single-Factor Test (Podsakoff et al., 2003), the Partial Correlation Technique (Podsakoff
et al., 2003), the Multitrait-Multimethod Technique (Shadish, Cook, & Campbell, 2002),
the Correlation Marker Technique, the CFA Marker Technique Correlation Marker
Technique (Lindell & Whitney, 2001) and the Unmeasured Latent Marker technique
(Liang, Saraf, Hu, & Xue, 2007) all have limitations in their ability to detect common
method bias. For example, while the Unmeasured Latent Marker Construct common
method bias test was demonstrated for use in PLS by Liang et al., more recent evidence
has demonstrated that the technique does not detect common method bias in PLS
analyses (Chin, Thatcher, & Wright, 2012). In addition to the procedural remedies, the
suggest test of first-order latent variable correlations being less than 0.9 was met in both
103
the Pretest and Test (Bagozzi et al., 1991). Additionally, the Correlation Marker
Technique also demonstrated only a 2.87% shared variance, which is less than the 3%
threshold, on the Test data. However, the potential of common method bias exists as data
was collected using an identical method and during a common timeframe. To reduce the
potential of this bias, it is recommended that the scale be applied to a different population
and at a different time in the future (Podsakoff et al., 2003). Furthermore, while studies
utilizing self-reported measures are suspected to contain common method bias, recent
research has demonstrated that studies utilizing self-reported measures have been shown
to offset such bias by attenuating effects of other errors, such as measurement error
(Conway & Lance, 2010).
Due to these limitations, further refinement and validation of the measurement
items must be performed. Examples of suggested future research include the application
on real-world cases, as well as measuring the construct on business decision-makers who
utilize geospatial data and comparing these results to average business decision-makers.
Furthermore, the GRA scale could be employed to extend studies related to usability of
SDSS and GIS, as well as to empirically measure the effectiveness of various
geovisualization techniques. To establish nomological validity, a confirmatory study
should be conducted to empirically test the GRA construct within the context of a
literature-based research model.
Additionally, as suggested by MacKenzie et al. (2011), an essential part of the
scale development process is to cross-validate the scale. Furthermore, as suggested by
MacKenzie et al., the development of norms to facilitate the interpretation of a
measurement scale is essential. It is suggested that a scale be tested on various population
104
groups to determine the generalizability to distinct groups. Additionally, we suggest that
a cluster analysis be performed to determine the feasibility of utilizing the GRA
measurement scale to classify subjects by their GRA.
As this study was conducted in a general context, and since geospatial reasoning
has been found to be highly task dependent, the study will need to be replicated in a
variety of domains to assess the impact of task characteristics on geospatial reasoning
ability. Also, as the study was presented in the context of the impact of geospatial
reasoning on decision performance, the resulting construct and instrument will need to be
applied in a decision-making domain, so that the overall effectiveness of the scale for
predicting the outcome variable can be assessed.
Finally, additional research should be performed to assess the SPGS construct that
was eliminated from the measurement model. Particularly, since the SPGS appeared to
have strong content, convergent and discriminant validity as well as reliability. While
SPGS did not appear to be predictive of the GRA construct, the analysis performed
indicated high reliability for this construct. Perhaps, SPGS will play an important role in
moderating decision-performance, thus additional research is warranted.
Conclusion
There has been rapid growth of geospatial information collected by business
organizations and advanced tools have been developed to visualize such data for
decision-making. However, while there have been several attempts to clarify what
cognitive abilities are necessary to make such decisions, there have been conflicting
results.
105
IS research has generally only measured one dimension of spatial ability, instead
of testing for a spectrum of spatial abilities. As such, the IS scholarship lacks an
empirically-validated, multi-dimensional, measurement of GRA, thus a multi-step
measurement scale development process, based on the work of MacKenzie et al. (2011),
was utilized to develop a multi-dimensional scale with reliability and validity.
The result of this research project was the development of a 12-item scale to
measure the three proposed substrata of GRA: (1) self-perceived geospatial orientation
and navigation, (2) self-perceived geospatial memorization and recall, and (3) self-
perceived geospatial visualization. As such, this measurement scale addresses multiple
dimensions of GRA, a limitation of previous studies addressing spatial ability.
Furthermore, the scale focuses on the geographic context of spatial reasoning. Finally, the
scale was developed to assess psychometric responses regardless of the past-geospatial
experience of research subjects. We suggest that future research further the validity of
these substrata and the GRA construct.
Finally, it was revealed that SPGS is an independent construct that is different
from GRA, but one that may also have an impact on decision-performance. It was
suggested that future research expand on the importance of SPGS in business decision-
making using geospatial data.
106
CHAPTER IV
USER-ACCEPTANCE OF SPATIAL DECISION SUPPORT
SYSTEMS: APPLYING UTILITARIAN, HEDONIC AND
COGNITIVE MEASURES3
Abstract
Following the scale development of GRA, Chapter IV presents an extension of
the Technology Acceptance Model in the context of geospatial visualization and the use
of online mapping services. The purpose of this is to further validate the GRA construct
and scale through the establishments of norms, external validity and nomological
validity.
SDSS have become an important tool for consumer and business evaluation of
geospatial information. This paper demonstrates that the user-acceptance of such tools is
influenced by utilitarian, hedonic and cognitive measures. Furthermore, the nomological
validity of geospatial reasoning ability is expanded through the use of well-established
constructs, including perceived enjoyment and perceived ease-of-use. Responses from
577 subjects were evaluated using a partial least squares analysis.
Introduction
Recent technological innovations related to collecting, analyzing and presenting
geospatial data have provided both consumer and business decision-makers with
advanced decision-support capabilities. For instance, the implementation of positioning
technologies, such as global satellite navigation systems and terrestrial navigation
capabilities provided by wireless network access points and mobile communication
3 A subsequent version of this chapter is under review at Information & Management.
107
networks, allows devices to easily determine their location in physical space.
Contemporary mobile devices, including tablets and smartphones, maximize the
capabilities of such positioning technologies by providing the capability to transmit
location information in real-time. In addition, analytic systems such as GIS and SDSS
have provided professionals and consumers with the capability to analyze geospatial
information more easily.
Consumers have benefited from the capabilities provided by online mapping
services, location-based services and web-based SDSS capabilities. Particularly, online
mapping services, such as Google Maps and Bing Maps, have experienced tremendous
usage growth in recent years (Google Maps, 2013; Bing Maps, 2013). Globally, over 100
million mobile users access Google Maps each month (Gundotra, 2010). In the United
States alone, about 48 million users accessed an online mapping service from a mobile
device, while nearly 94 million users accessed such services from a fixed Internet access
point (ComScore, 2011). Most online mapping services and location-based services
utilize geovisualization to present geospatial information to the user. Thus, an
understanding of user perceptions and attitudes toward geovisualization are essential for
future industry development and academic research.
In addition to a direct impact on consumers, geovisualization also became an
important presentation technique for conveying complex geospatial relationships
contained in enterprise data to business decision-makers. As over 75 percent of all
business data contains geographic information and 80 percent of all business decisions
utilize geographic data (Mennecke, 1997; Tonkin, 1994), it has become essential to
develop geovisualization tools and techniques to maximize decision performance for
108
business leaders. Furthermore, location analytics, which include social, emotional,
geographic and physical indicators, are providing business decision-makers with even
more advanced insights into their geospatial data (Ferguson, 2012). Often such
aggregated data is presented in a geovisualized presentation format designed to allow
decision-makers to easily identify complex geospatial interactions. Likened to a
consumer context, a better understanding of geovisualization in the business context is
essential.
A lack of research addressing user perceptions and attitudes regarding
geovisualization provided the motivation for this study. Particularly, this study measured
utilitarian, hedonic and cognitive user perceptions to better understand user acceptance of
geovisualization. The utilitarian perceptions were measured through the inclusion of the
perceived usefulness and perceived ease-of-use constructs. The hedonic perceptions were
provided by measures of the perceived enjoyment construct. Finally, cognitive
perceptions were measured through the implementation of the geospatial reasoning
ability construct. These constructs, along with their appropriate theoretical background,
are described next.
Perceived Enjoyment: Hedonic Measures
Numerous researchers have found that hedonic measures, like fun or enjoyment,
are relevant for predicting user acceptance and use of information technologies (e.g.
Brown & Venkatesh, 2005; Childers, Carr, Peck & Carsons, 2002; Gerow, Ayyagari,
Thatcher & Roth, 2013; Thong, Hong & Tam, 2006; van der Heijden, 2004). Studies
have found that hedonic aspects of systems are improved in systems that include
appealing visual layouts like graphics and colors (Ives, 1982; Klein, Moon, & Picard,
109
2002; van der Heijden, 2004). The inherent visual nature of geovisualization systems
suggests that understanding the hedonic attributes might be especially important to
understanding user-acceptance of these systems. For this study, the hedonic measures
included in the perceived enjoyment (PE) construct developed by Lee, Fiore and Kim
(2006) were utilized.
Technology Acceptance Model: User-Acceptance and Utilitarian Measures
Davis, Bagozzi and Warshaw (1989) adapted Fishbein and Ajzen’s (1975) Theory
of Reasoned Action to develop the Technology Acceptance Model (TAM). The Theory
of Reasoned Action posits that beliefs influence attitudes. The initial TAM assumed that
utilitarian views about perceived usefulness (PU) and perceived ease-of-use (PEOU)
influence technology acceptance. The TAM has been empirically applied to better
understand the acceptance of various technologies, such as online shopping (Gefen,
Karahanna & Straub, 2003), enterprise technologies (Amoako-Gyampah & Salam, 2004;
Amoako-Gyampah, 2007) and even telemedicine (Hu, Chau, Sheng & Tam, 1999).
However, TAM has not been applied to the concept of geovisualization or technologies
that rely on geovisualization. This study will address this gap and provide an initial
understanding of which factors influence the user acceptance of SDSS. An
understanding, of the utilitarian perceptions of SDSS will allow industry to begin the
development of more effective geovisualization techniques.
In addition to measuring PU and PEOU, TAM traditionally also includes
measures of attitude toward a technology (A) and the behavioral intent to use a
technology (BI), which are included as well.
110
Geospatial Reasoning Ability: Cognitive Measures
Erskine and Gregg’s (2011) measure of geospatial reasoning ability (GRA) was
developed to asses an individual’s perceived cognitive ability to make decisions using
geospatial information. Geospatial reasoning ability is measured using a 12-item scale
encompassing the three substrata of geospatial reasoning: (1) self-perceived geospatial
orientation and navigation, (2) self-perceived geospatial memorization and recall, and (3)
self-perceived geospatial visualization. While most prior information systems research
evaluated spatial ability using only one or two dimensions of spatial ability, geospatial
reasoning ability evaluates a spectrum of spatial abilities. Furthermore, the measurement
scale focuses on the geographic context of spatial reasoning. Additionally, the scale was
developed to assess psychometric responses, regardless of prior geospatial experience,
making it appropriate for expert and non-expert subjects.
This research project was established to investigate the following questions: (1)
Do cognitive measures, such as geospatial reasoning ability, impact utilitarian
perceptions of SDSS? (2) Do cognitive measures, such as geospatial reasoning ability,
impact hedonic perceptions of SDSS? (3) Do utilitarian perceptions, such as perceived
ease-of-use of SDSS, impact attitude toward SDSS and behavioral intent of using SDSS?
(4) Do hedonic perceptions, such as perceived enjoyment of SDSS, impact attitude
toward SDSS and behavioral intent of using SDSS? In addition to answering these
research questions, another goal of this study was the initial development of a
nomological network via well-established constructs for the relatively new geospatial
reasoning ability construct.
111
In the following section a research model and related hypotheses are introduced.
This is followed by an explanation of the research methodology applied in this study.
Then, an analysis of the dataset is presented. Finally, a discussion of findings, disclosure
of limitations, suggestions for future research and a conclusion are presented.
Research Model
This section introduces the theoretical lens, a proposed research model and
hypotheses developed to better understand the user-acceptance of geovisualization
through online mapping services. This study will expand on Davis et al.’s (1989) TAM, a
widely cited theory of technology acceptance based on Fishbein and Ajzen’s (1975)
Theory of Reasoned Action. This extension of TAM combines utilitarian, hedonic and
cognitive perceptions to determine attitudes and the extent of behavioral intent.
Furthermore, this research model utilizes TAM to better understand user acceptance
within the geovisualization domain, a technique which conveys complex geospatial
relationships and is utilized in various contemporary technologies including online
mapping services and advanced GIS and SDSS technologies. Figure 8 introduces the
proposed research model and its associated hypotheses visually, which are then explained
thereafter.
112
Figure 8. Proposed Research Model.
Prior research has shown that technologies such as location-based services and
virtual reality platforms, which require spatial reasoning abilities, also have an effect on
the hedonic perceptions of subjects. For instance, Winn, Hoffman, Hollander, Osberg,
Rose and Char (1997) highlight the influence of spatial ability on enjoyment and
presence within virtual environments. Ho (2010) found a significant relationship between
the use of location-based services and perceived enjoyment. In addition, Strohecker
(2000) notes that subjects who utilized a graphical software tool that allowed for the
simulation of scale, perception and frame of reference, often expressed enjoyment. In
their study, Hwang, Jung and Kim (2006) learned of a significant relationship between
the use of hand-held, motion-based virtual reality platforms and enjoyment. While the
aforementioned studies demonstrate effects on hedonic perceptions, other information
technologies such as customer relationship management tools did not appear to have a
statistically significant impact on enjoyment (Al-Momani & Noor, 2009). Thus, the
author posits:
H1: Geospatial Reasoning Ability (GRA) positively influences Perceived
Enjoyment (PE)
113
In addition to influencing hedonic measures, geospatial reasoning ability is likely
to impact utilitarian measures. While geospatial reasoning ability has not been used as a
direct indicator of PEOU or PU in previous studies, several existing studies provide
evidence of a significant positive relationship between geospatial reasoning ability and
perceived ease-of-use, and also between geospatial reasoning ability and perceived
usefulness (Arning & Ziefle, 2007; Campbell, 2011; Gilbert, Lee-Kelley & Barton,
2003). In their study, Arning and Ziefle determined that spatial ability had a significant
effect on the performance of using a personal digital assistant (PDA) for information
retrieval. In two experiments, spatial ability measured by spatial visualization and spatial
orientation predicted effectiveness and efficiency, which are measures of usability,
respectively (Campbell, 2011). Furthermore, it is suggested that the implications of
spatial ability on psychological attitude and behavior should not be ignored (Gilbert et al.,
2003). Thus, the author posits:
H2: Geospatial Reasoning Ability (GRA) positively influences Perceived Ease-of-
Use (PEOU)
H3: Geospatial Reasoning Ability (GRA) positively influences Perceived
Usefulness (PU)
The positive and significant effect of PEOU on PE has been demonstrated in
numerous empirical studies (e.g. Balog & Pribeanu, 2010; Chesney, 2006; van der
Heijden, 2004). van der Heijden identified a significant effect of perceived ease-of-use on
perceived enjoyment in an evaluation of a cinema website, which would be considered a
highly hedonic information system. Chesney discovered a positive relationship between
perceived ease-of-use and perceived enjoyment in a study investigating Lego Mindstorms
114
development environments. Balog and Pribeanu also discovered a significant relationship
between perceived ease-of-use and perceived enjoyment in their study of augmented
reality teaching platforms. Based on the numerous studies that identified a significant and
positive relation between perceived ease-of-use and perceived enjoyment, the author
posits:
H4: Perceived Ease-of-Use (PEOU) positively influences Perceived Enjoyment
(PE)
PEOU has been found to have a significant relationship with attitude in numerous
empirical studies (e.g. Chen, Gillenson & Sherrell, 2002; Chen & Tan, 2004; Lin & Lu,
2000; O’Cass & Fenech, 2003). In their study, Chen et al. (2002) discovered that PEOU
use is a key antecedent of A. Chen and Tan (2004) reported a significant relationship
between PEOU on A. In their study of websites, Lin and Lu discovered that PEOU had a
significant relationship with A. In another empirical study, O’Class and Fenech
discovered that PEOU had a significant positive effect on web retail A. Vijayasarathy
(2004) also added that significant relationships exist between the PEOU and A of online
shopping. Furthermore, Davis et al. (1989) stated that the combined PEOU and PU
determine the attitude toward a system. Thus, we posit:
H5: Perceived Ease-of-Use (PEOU) positively influences Attitude (A)
Additionally, numerous studies have demonstrated a relationship between PEOU
and PU (e.g., Davis, 1993; Subramanian, 1994). Furthermore, Legris, Ingham and
Collerette (2003) cited twenty-one studies that indicated a positive relationship between
PEOU and PU. It is clear, based on prior research, that PEOU influences perceived
115
usability across a wide variety of domains (e.g. Davis, 1989, 1993; Davis et al., 1989).
Thus, we posit that:
H6: Perceived Ease-of-Use (PEOU) positively influences Perceived Usefulness
(PU)
Hedonic aspects of technology adoption have been considered in numerous
technology adoption studies (e.g., Davis, Bagozzi & Warshaw, 1992; van der Heijden,
2004). Several papers specifically indicate relationships between PE and BI within the
context of mobile and online mapping services (e.g., Abad, Díaz & Vigo, 2010; Ho,
2010; Novak & Schmidt, 2009; Verkasalo, López-Nicolás, Molina-Castillo & Bouwman,
2010). Novak and Schmidt discovered a significant relationship between hedonic
measures and the use of a large collaborative display that provided information
contextualized on a map in a travel advisory scenario. Ho found a significant relationship
between the use of location-based services and PE. Abad et al. found significant
relationships between PE and BI in an outdoor content generation exercises that utilized
Google Maps along with other mobile device features. Verkasalo et al. also found that BI
was impacted by PE of mobile map applications.
Based on the importance of hedonic aspects, such as PE, on user acceptance of
technologies, we posit:
H7: Perceived Enjoyment (PE) positively influences Attitude (A)
PU has been found to significantly influence A in numerous empirical studies
(e.g., Chen et al., 2002; Chen & Tan, 2004; Lin & Lu, 2000; O’Cass & Fenech, 2003;
Vijayasarathy, 2004). In their study regarding online retailers, Lee et al. (2006)
discovered a significant relationship between the PU of online shopping and A toward an
116
online retailer. Chen et al. (2002) discovered a significant positive relationship between
the PU of an online store and a consumer’s A toward the store. In Chen and Tan (2004), a
significant relationship between PU and A was reported. As reported in their extension of
the TAM, Gefen and Straub (1997) noted that PU and PEOU affect attitude toward use.
In their study of websites, Lin and Lu (2000) discovered that perceived usefulness
had a significant relationship with attitude as measured by preferences of a web site. In
another empirical study, O’Class and Fenech (2003) discovered that perceived usefulness
had a positive effect on attitude of web retail. Vijayasarathy (2004) also discovered a
significant relationship between perceived usefulness and attitude of online shopping.
Furthermore, Davis et al. (1989) stated that the combined perceived ease-of-use and
perceived usefulness determine the attitude towards a system. Thus, we posit:
H8: Perceived Usefulness (PU) positively influences Attitude (A)
While the early version of the TAM did not include attitude (Davis et al., 1989),
later works (Davis, 1993; Mathieson, 1991) included attitude as an antecedent toward
behavioral intent. Furthermore, Davis (1993) reported that the overall attitude of a
technology is a significant determinant of whether someone will use it. Liker and Sindi
(1997) reported that positive attitudes of a technology would result in usage of a
technology, while negative attitudes will result in rejection of a technology. Positive
attitude, positive/negative feelings, and evaluated belief toward an object, are all related
to behavioral intent (Karahanna, Straub & Chervany, 1999). Furthermore, the relationship
between attitude and behavioral intent has been widely tested in the literature (e.g.,
Anderson & Agarwal, 2010; Herath & Rao, 2009; Venkatesh, Morris, Davis & Davis,
117
2003). As a result, we expect that attitude regarding location-aware mobile applications is
positively related to intentions of using SDSS. Thus, we posit that
H9: Attitude (A) positively influences Behavioral Intent (BI)
Next, the research methodology used to test the aforementioned hypotheses is
presented.
Research Methodology
To assess the above research model, an online survey instrument comprised of
measurement items adapted from prior research was utilized. Next, subjects who were
familiar with online mapping service were solicited to complete the survey. A paragraph
explaining geovisualization and online mapping services was provided, and only subjects
who had previously used online mapping services were allowed to complete the survey.
Research Sample
Subjects who were familiar with online mapping services were solicited to
complete a survey. All subjects were provided a paragraph explaining geovisualization
and online mapping services and then asked a screening question to determine if they had
ever used an online mapping service. Only subjects who had previously used a SDSS and
who were over the age of 18 were permitted to participate.
Five-hundred and seventy-seven usable responses were obtained from subjects
over a six-month period. Numerous techniques were utilized to contact potential subjects,
including e-mail, social network services and personal contact. The goal of the
recruitment process was to obtain subjects from a wide variety of backgrounds. One
limitation of the recruitment method was that it provided no meaningful way to
demonstrate response rate.
118
Research Instrument
The instrument first explained some basic concepts of geovisualization, geospatial
data, etc. Upon learning about these concepts, demographic information was collected
from each subject, including age, gender, education, and cultural background. Next, the
participants completed measurement items for hedonic, utilitarian, cognitive, attitudes
and behavioral intent. The utilized instruments are presented in Tables 41-46. The survey
instrument was distributed online using SurveyMonkey.
Several studies have included measures of hedonic perceptions in relationship to
information technologies. Table 41 lists the six PE measurement items utilized in this
study, which were adapted from Lee et al. (2006).
Table 41. Perceived Enjoyment Measurement Items.
Measurement Item Source/Adapted From
PE1: Viewing spatial information presented in a
geovisualized format would be entertaining.
Lee et al., 2006
PE2: Viewing spatial information presented in a
geovisualized format would be enjoyable.
Lee et al., 2006
PE3: Viewing spatial information presented in a
geovisualized format would be interesting.
Lee et al., 2006
PE4: Viewing spatial information presented in a
geovisualized format would be fun.
Lee et al., 2006
PE5: Viewing spatial information presented in a
geovisualized format would be exciting
Lee et al., 2006
PE6: Viewing spatial information presented in a
geovisualized format would be appealing.
Lee et al., 2006
Two established constructs were used to measure utilitarian perception of
geovisualization. The first of these constructs, PU, consisted of six measurement items
derived from Davis (1989), Davis et al. (1989), Venkatesh and Davis (2000), van der
119
Heijden (2004), and Fagan, Neill and Wooldridge (2008). These measurement items are
shown in Table 42.
Table 42. Perceived Usefulness Measurement Items.
Measurement Item Source/Adapted From
PU1: Viewing spatial information presented in a
geovisualized format would improve my decision-
making productivity.
Davis, 1989; Davis et
al., 1989; Fagan et al.,
2008; Venkatesh &
Davis, 2000
PU2: Viewing spatial information presented in a
geovisualized format would enhance my effectiveness
in decision-making.
Fagan et al., 2008;
Venkatesh & Davis,
2000
PU3: Viewing spatial information presented in a
geovisualized format would be useful.
Davis, 1989; Davis et
al., 1989; Fagan et al.,
2008; Venkatesh &
Davis, 2000
PU4: Viewing spatial information presented in a
geovisualized format would increase my decision-
making performance.
Davis, 1989; Davis et
al., 1989; Fagan et al.,
2008; Venkatesh &
Davis, 2000
PU5: Viewing spatial information presented in a
geovisualized format would make my decision-making
easier.
Davis, 1989; Davis et
al., 1989
PU6: Viewing spatial information presented in a
geovisualized format would allow me to be more
informed.
van der Heijden, 2004
The second of these constructs, PEOU, consists of six measurement items derived
from Davis (1989), Davis et al. (1989), Venkatesh and Davis (2000), and Fagan et al.
(2008). These measurement items are shown in Table 43.
120
Table 43. Perceived Ease-of-Use Measurement Items.
Measurement Item Source/Adapted From
PEOU1: Learning to operate a system that
geovisualizes information would be easy for me.
Davis, 1989; Davis et
al., 1989; Fagan et al.
2008
PEOU2: I would find it difficult to get a system that
geovisualizes information to do what I want it to.
Davis, 1989; Davis et
al., 1989
PEOU3: My interaction with a system that
geovisualizes information would be clear and
understandable.
Davis, 1989; Davis et
al., 1989; Venkatesh &
Davis, 2000
PEOU4: I would find a system that geovisualizes
information to be flexible to interact with.
Davis, 1989; Davis et
al., 1989
PEOU5: It would be easy for me to become skillful at
using a system that geovisualizes information.
Davis, 1989; Davis et
al., 1989
PEOU6: I would find a system that geovisualizes
information easy to use.
Davis, 1989; Davis et
al., 1989; Fagan et al.,
2008; Venkatesh &
Davis, 2000
The attitude construct consists of six measurement items that were adapted from
Lee et al. (2006). These items are shown in Table 44.
Table 44. Attitude Measurement Items.
Measurement Item Source/Adapted From
A1: Viewing spatial information using geovisualization
would be good.
Lee et al., 2006
A2: Viewing spatial information using geovisualization
would be superior.
Lee et al., 2006
A3: Viewing spatial information using geovisualization
would be pleasant.
Lee et al., 2006
A4: Viewing spatial information using geovisualization
would be excellent.
Lee et al., 2006
A5: Viewing spatial information using geovisualization
would be interesting.
Lee et al., 2006
A6: Viewing spatial information using geovisualization
would be worthwhile.
Lee et al., 2006
121
The BI construct consists of six measurement items that were adapted from
Venkatesh and Davis (2000). These items are shown in Table 45.
Table 45. Behavioral Intent Measurement Items.
Measurement Item Source/Adapted From
BI1: If I have access to a geovisualization technology I
intend to use it.
Venkatesh & Davis,
2000
BI2: I intend to utilize a geovisualization technology. Venkatesh & Davis,
2000
BI3: Assuming I have access to a geovisualization
technology, I intend to use it.
Venkatesh & Davis,
2000
BI4: Given that I have access to a geovisualization
technology, I predict that I would use it.
Venkatesh & Davis,
2000
BI5: I predict that I would utilize a geovisualization
technology.
Venkatesh & Davis,
2000
BI6: I plan to utilize a geovisualization technology. Venkatesh & Davis,
2000
The GRA construct consists of twelve items sourced from Erskine and Gregg
(2011). These items are shown in Table 46.
122
Table 46. Geospatial Reasoning Ability Measurement Items.
Measurement Item Source/Adapted From
GRA1: I can usually remember a new route after I have
traveled it only once.
Erskine & Gregg, 2011
GRA2: I am good at giving driving directions from
memory.
Erskine & Gregg, 2011
GRA3: After studying a map, I can often follow the
route without needing to look back at the map.
Erskine & Gregg, 2011
GRA4: I am good at giving walking directions from
memory.
Erskine & Gregg, 2011
GRA5: In most circumstances, I feel that I could
quickly determine where I am based on my
surroundings.
Erskine & Gregg, 2011
GRA6: I have a great sense of direction. Erskine & Gregg, 2011
GRA7: I feel that I can easily orientate myself in a new
place.
Erskine & Gregg, 2011
GRA8: I rarely get lost. Erskine & Gregg, 2011
GRA9: I can visualize geographic locations. Erskine & Gregg, 2011
GRA10: I can visualize a place from information that is
provided by a map without having been there.
Erskine & Gregg, 2011
GRA11: I can visualize a place from a map. Erskine & Gregg, 2011
GRA12: While reading written walking directions, I
often form a mental image of the walk.
Erskine & Gregg, 2011
Upon completion of the data collection, an analysis was performed using
SmartPLS (Ringle et al., 2005).
Analysis
The measurement and structural models of the proposed research were assessed
by a partial least squares analysis performed using Smart PLS, which analyzed the
measurement (outer) and structural (inner) model. The demographic statistics of the
sample can be found in Table 47.
123
Table 47. Descriptive Statistics of Demographic Variables.
Question Variables Percentage
Age 18-25 31.4%
26-35 32.4%
36-45 11.8%
45-55 12.5%
56-65 8.0%
66+ 4.0%
Gender Female 55.6%
Male 44.4%
Education 2 Year/Associate Degree 25.5%
4 Year/Bachelor Degree 30.5%
Doctor/JD/PhD 2.8%
Elementary/Middle School 0.9%
High School 30.3%
Master Degree 10.1%
Profession Business Professional 42.2%
Student 35.9%
Geospatial Professional 0.5%
Other 18.4.0%
Culture African 3.8%
Asian 11.3%
Australian 0.2%
European 35.9%
Middle Eastern 4.2%
North American 43.5%
South American 1.2%
The measurement model explains the relationship between each of the constructs
and its corresponding measurement items. We assessed the measurement model by
evaluating the item reliability, internal consistency and discriminant validity. Establishing
an acceptable measurement model allowed the structural model to be tested.
The first task involved assessing the reliability of the measurement items. The
first step of this task included an evaluation of each measurement item’s loading and the
removal of any item with a loading of less than 0.5. All items except PEOU2 were
deemed acceptable. A detailed examination of the data set revealed that the results of
124
PEOU2 correctly inversely mirrored similar responses, to a lesser extent, and that most
responses were neutral. We feel that the use of a negatively worded question may have
fundamentally changed the meaning of the item. For example, a person may not find a
tool easy to use, but that does not necessarily mean that the tool is inherently difficult to
them. The use of negatively and positively worded items in a research study has been a
topic in prior research with mixed results (e.g. Ibrahim, 2001; Mook, Kleijn & van der
Ploeg, 1992; Schmitt & Stults, 1985; Spector, Van Katwyk, Brannick & Chen, 1997).
Thus, PEOU2 was removed from the analysis.
The reliability of the measurement items is shown in Table 48. As composite
reliability and Cronbach’s (1951) alpha values of greater than 0.7 are deemed acceptable,
all values appear excellent, which demonstrates measurement item reliability.
Table 48. Item Reliability.
Construct Composite Reliability Cronbach’s Alpha
GRA 0.9457 0.9371
PE 0.9652 0.9567
PU 0.9606 0.9506
PEOU 0.9452 0.9274
A 0.9450 0.9301
BI 0.9667 0.9586
Generally, acceptable thresholds have stated that items should load at greater than
0.7 on their own constructs and cross-loadings should be less than 0.5 on other
constructs. Other “rules-of-thumb” include that item cross-loadings should differ by more
than 0.2, and that all correlations between factors should not exceed 0.7. While all
measurement items except GV10 exceed the 0.7 threshold, many cross-loadings were
higher than the suggested 0.5 minimum (Hair et al., 2010). However, each item loaded
more highly on its own construct and constructs appear to share more variance with their
125
respective measures. Table 49 demonstrates item loading and cross loading of the outer
model. Per Vinzi, Chin, Henseler, and Wang (2010), the loading and cross-loading values
can be misleading unless they are squared, thus a second table (Table 50) which
demonstrates squared loadings and cross-loadings is presented.
126
Table 49. Item Loadings and Cross Loadings, highest loadings shown in bold.
A BI GRA PE PEOU PU
A1 0.8625 0.6435 0.3518 0.5812 0.6259 0.6814
A2 0.8321 0.6045 0.3142 0.5357 0.5623 0.6353
A3 0.8683 0.6347 0.3287 0.6637 0.5827 0.6036
A4 0.8786 0.6312 0.3618 0.6511 0.5958 0.6658
A5 0.8657 0.6041 0.3494 0.6953 0.5741 0.6652
A6 0.8576 0.6603 0.3534 0.6111 0.5984 0.7002
BI1 0.6197 0.8664 0.3511 0.5083 0.5820 0.5841
BI2 0.6461 0.9165 0.3944 0.5084 0.6244 0.5853
BI3 0.6796 0.9282 0.3729 0.5735 0.6073 0.6174
BI4 0.6687 0.9011 0.3526 0.5129 0.5869 0.5762
BI5 0.7111 0.9354 0.3922 0.5607 0.6450 0.6278
BI6 0.6676 0.9131 0.3996 0.5214 0.6233 0.5964
GMR1 0.2755 0.2961 0.7826 0.3564 0.3900 0.2448
GMR2 0.2332 0.2798 0.7779 0.3439 0.3661 0.2462
GMR3 0.3223 0.3413 0.8141 0.3841 0.4613 0.3286
GMR4 0.3077 0.3424 0.7790 0.3861 0.3645 0.3441
GON10 0.2624 0.2853 0.7580 0.3209 0.3823 0.2548
GON3 0.3366 0.3238 0.7496 0.3371 0.4144 0.2789
GON4 0.3054 0.3146 0.8227 0.3382 0.4172 0.2569
GON5 0.2485 0.2790 0.7874 0.2646 0.3809 0.2246
GV1 0.3733 0.3248 0.7941 0.4276 0.4453 0.3833
GV10 0.3138 0.3160 0.5977 0.3524 0.3704 0.3272
GV2 0.3080 0.3242 0.7879 0.3678 0.4276 0.2966
GV3 0.3544 0.3705 0.7696 0.4061 0.4610 0.3422
PE1 0.6280 0.5204 0.4377 0.9042 0.5120 0.5631
PE2 0.6456 0.5306 0.4568 0.9416 0.5241 0.5819
PE3 0.6713 0.5459 0.4410 0.8976 0.5241 0.6216
PE4 0.6610 0.5344 0.4091 0.9210 0.5164 0.5735
PE5 0.6482 0.4987 0.4149 0.8962 0.4783 0.5771
PE6 0.6839 0.5430 0.3926 0.8788 0.5229 0.6561
PEOU1 0.5601 0.5773 0.4973 0.5259 0.8479 0.5223
PEOU3 0.6111 0.6086 0.4824 0.4852 0.9008 0.5377
PEOU4 0.6339 0.5892 0.4311 0.4982 0.8635 0.5263
PEOU5 0.5933 0.5950 0.4832 0.4892 0.9032 0.5211
PEOU6 0.6185 0.5873 0.4492 0.4934 0.8858 0.5489
PU1 0.6991 0.5871 0.3454 0.5807 0.5595 0.9195
PU2 0.6877 0.5697 0.3569 0.5909 0.5395 0.9252
PU3 0.6630 0.5821 0.3330 0.5831 0.5083 0.8663
PU4 0.6723 0.5809 0.3466 0.5811 0.5293 0.9042
PU5 0.6828 0.5817 0.3193 0.5795 0.5395 0.9062
PU6 0.7058 0.6266 0.3807 0.6161 0.5641 0.8513
127
Table 50. Square Root of Loadings and Cross-Loadings, highest loadings shown in
bold.
A BI GRA PE PEOU PU
A1 0.9287 0.8022 0.5931 0.7624 0.7911 0.8255
A2 0.9122 0.7775 0.5605 0.3719 0.7499 0.7971
A3 0.9318 0.7967 0.5733 0.8147 0.7633 0.7769
A4 0.9373 0.7945 0.6015 0.8069 0.7719 0.8160
A5 0.9304 0.7772 0.5911 0.8338 0.7577 0.8156
A6 0.9261 0.8126 0.5945 0.7817 0.7736 0.8368
BI1 0.7872 0.9308 0.5925 0.7130 0.7629 0.7643
BI2 0.8038 0.9573 0.6280 0.7130 0.7902 0.7650
BI3 0.8244 0.9634 0.6107 0.7573 0.7793 0.7857
BI4 0.8177 0.9493 0.5938 0.7162 0.7661 0.7591
BI5 0.8433 0.9672 0.6263 0.7488 0.8031 0.7923
BI6 0.8171 0.9556 0.6321 0.7221 0.7895 0.7723
GMR1 0.5249 0.5442 0.8846 0.5970 0.6245 0.4948
GMR2 0.4829 0.5290 0.8820 0.5864 0.6051 0.4962
GMR3 0.5677 0.5842 0.9023 0.6198 0.6792 0.5732
GMR4 0.5547 0.5851 0.8826 0.6214 0.6037 0.5866
GON10 0.5122 0.5341 0.8706 0.5665 0.6183 0.5048
GON3 0.5802 0.5690 0.8658 0.5806 0.6437 0.5281
GON4 0.5526 0.5609 0.9070 0.5815 0.6459 0.5069
GON5 0.4985 0.5282 0.8874 0.5144 0.6172 0.4739
GV1 0.6110 0.5699 0.8911 0.6539 0.6673 0.6191
GV10 0.5602 0.5621 0.7731 0.5936 0.6086 0.5720
GV2 0.5550 0.5694 0.8876 0.6065 0.6539 0.5446
GV3 0.5953 0.6087 0.8773 0.6373 0.6790 0.5850
PE1 0.7925 0.7214 0.6616 0.9509 0.7155 0.7504
PE2 0.8035 0.7284 0.6759 0.9704 0.7239 0.7628
PE3 0.8193 0.7389 0.6641 0.9474 0.7239 0.7884
PE4 0.8130 0.7310 0.6396 0.9597 0.7186 0.7573
PE5 0.8051 0.7062 0.6441 0.9467 0.6916 0.7597
PE6 0.8270 0.7369 0.6266 0.9374 0.7231 0.8100
PEOU1 0.7484 0.7598 0.7052 0.7252 0.9208 0.7227
PEOU3 0.7817 0.7801 0.6946 0.6966 0.9491 0.7333
PEOU4 0.7962 0.7676 0.6566 0.7058 0.9292 0.7255
PEOU5 0.7703 0.7714 0.6951 0.6994 0.9504 0.7219
PEOU6 0.8764 0.7664 0.6702 0.7024 0.9412 0.7409
PU1 0.8361 0.7662 0.5877 0.7620 0.7480 0.9589
PU2 0.8293 0.7548 0.5974 0.7687 0.7345 0.9619
PU3 0.8142 0.7630 0.5771 0.7636 0.7130 0.9308
PU4 0.8199 0.7622 0.5887 0.7623 0.7275 0.9509
PU5 0.8263 0.7627 0.5651 0.7612 0.7345 0.9519
PU6 0.8401 0.7916 0.6170 0.7849 0.7511 0.9227
128
Next, discriminant validity was examined. Discriminant validity demonstrates the
extent to which one construct or variable is distinct from another (Hair et al., 2010).
While loadings and cross-loadings can be used to establish discriminant validity, such
validity can also be examined by comparing inter-construct correlations with the square
root of the average variance extracted (AVE). In such an examination, the square root of
the AVE should be equal or greater than any of the inter-construct correlations. See Table
51 for a comparison of the inter-construct correlations and the square root of each
construct’s AVE.
Table 51. Inter-Construct Correlations and Square Root of AVE.
GRA PE PU PEOU A BI Square Root of
AVE
GRA - 0.4692 0.3879 0.5323 0.3991 0.4144 0.7703
PE 0.4692 - 0.6575 0.5661 0.7243 0.5837 0.9068
PU 0.3879 0.6575 - 0.6036 0.7655 0.7655 0.8959
PEOU 0.5323 0.5661 0.6036 - 0.6855 0.6719 0.8805
A 0.3991 0.7243 0.7655 0.6855 - 0.7318 0.8609
BI 0.4144 0.5837 0.6570 0.6719 0.7318 - 0.9104
Next, validity is examined using the AVE and latent variable correlations. When a
construct AVE exceeds a threshold of 0.5, convergent validity can be established (Fornell
& Larcker, 1981). Table 52 demonstrates that the construct AVE values exceed this
threshold.
129
Table 52. Construct AVE.
Construct AVE
Geospatial Reasoning Ability (GRA) 0.5934
Perceived Enjoyment (PE) 0.8223
Perceived Usefulness (PU) 0.8026
Perceived Ease-of-Use (PEOU) 0.7753
Attitude (A) 0.7412
Behavioral Intent (BI) 0.8288
Upon completion of the assessment of the measurement model an analysis of the
structural model was performed. To accomplish such an analysis, each of the inner model
relationships was analyzed. The first step of analyzing the structural model was to
perform an assessment of the effects as shown in Figure 9.
*significant at alpha of 0.001, ^ significant at alpha of 0.05 Figure 9. Initial Nomological Network of GRA, along with Path Coefficients and R
Square Values.
The relationships between GRA and PE, PEOU and PU were positive with path
coefficients of 0.234, 0.532 and 0.093, respectively. These associations are consistent
with Hypotheses H1, H2 and H3. Furthermore, the GRA construct accounts for 28.3% of
the variance of PEOU in the structural model.
The relationships between PEOU and PE, A, and PU were positive with path
coefficients of 0.441, 0.269 and 0.554, respectively. These relationships are consistent
130
with Hypotheses H4, H5 and H6. Furthermore, the combined effects of GRA and PEOU
account for 36.0% of the variance of PE in the structural model, and the combined effects
of GRA and PEOU account for 37.1% of the variance of PU.
The relationships between PE and A, as well as PU and A, were positive with
path coefficients of 0.309 and 0.400, respectively. These relationships are consistent with
Hypotheses H7 and H8. Furthermore, the combined effects of PE, PEOU and PU account
for 71.4% of the variance found in A.
The relationship of A and BI was positive with a path coefficient of 0.732, which
is consistent with hypothesis H9. Furthermore, the effect of A accounts for 53.6% of the
variance of BI.
Finally, the significance of each of the construct relationships was tested using the
Smart PLS bootstrap re-sampling procedure with 200 re-samples and using 577 cases.
The purpose of the test statistic was to determine the significance relationship seen in the
PLS model. It is recommended that t-statistic values should be greater than 1.96 for an
alpha 0.05 or 2.56 for an alpha of 0.001 (Gefen & Straub, 2005). The alpha of 0.001
threshold was met by all relationships except GRA PU, which only met the alpha of
0.05 threshold. See Table 53 for the path coefficient table that includes the t-statistics
results.
131
Table 53. Path Coefficients.
Relationship Original
Sample
Sample
Mean
Standard
Deviation
Standard
Error
T
Statistics
A -> BI 0.7318 0.7344 0.0215 0.0215 34.1083
GRA -> PE 0.2342 0.2394 0.0488 0.0488 4.7985
GRA -> PEOU 0.5323 0.5386 0.0359 0.0359 14.8165
GRA -> PU 0.0929 0.0950 0.0471 0.0471 1.9728
PE -> A 0.3090 0.3109 0.0382 0.0382 8.0914
PEOU -> A 0.2693 0.2725 0.0440 0.0440 6.1130
PEOU -> PE 0.4415 0.4405 0.0480 0.0480 9.1947
PEOU -> PU 0.5542 0.5569 0.0484 0.0484 11.4416
PU -> A 0.3998 0.3976 0.0430 0.0430 9.2905
Discussion
The results of the analysis reveal conclude that all hypotheses are supported.
These results are summarized in Table 54.
Table 54. Hypothesis Tests.
Hypothesis Findings
H1: Geospatial Reasoning Ability positively influences
Perceived Enjoyment
Significant/Positive*
H2: Geospatial Reasoning Ability positively influences
Perceived Ease-of-Use
Significant/Positive*
H3: Geospatial Reasoning Ability positively influences
Perceived Usefulness
Significant/Positive^
H4: Perceived Ease-of-Use positively influences
Perceived Enjoyment
Significant/Positive*
H5: Perceived Ease-of-Use positively influences Attitude Significant/Positive*
H6: Perceived Ease-of-Use positively influences
Perceived Usefulness
Significant/Positive*
H7: Perceived Enjoyment positively influences Attitude Significant/Positive*
H8: Perceived Usefulness positively influences Attitude Significant/Positive*
H9: Attitude positively influences Behavioral Intent Significant/Positive*
* significant at alpha of 0.001
^ significant at alpha of 0.05
132
Based on the results of the hypotheses tests, an updated research model (Figure 9)
is presented. Additionally, this model serves as an initial nomological network of the
geospatial reasoning ability construct.
This study examines the impact of geospatial reasoning ability on an individual’s
perceived enjoyment, perceived ease of use and perceived usefulness associated with
using an online mapping tool. Our hypotheses were all supported, as shown in Table 54.
This suggests that an individual’s underlying geospatial reasoning abilities have a
significant impact on their hedonic and utilitarian perception. This is consistent with prior
research investigating technology acceptance in other domains that have found that an
individual’s traits, such as mood, self-efficacy and personality traits strongly influence
user perception of technological tools (e.g., Devaraj, Easley & Crant, 2008; Djamasbi,
Strong & Dishaw, 2010; Luo, Li Zhang & Shim, 2010).
Consistent with the TAM, this research suggests that utilitarian and hedonic
measures are indicators of technology acceptance (Davis et al., 1992). However, this
model is extended by evaluating the impact of geospatial cognition on technology
acceptance, as measured by GRA. This has important implications in today's environment
where geospatial data is increasingly being added to commonly used devices, such as
smartphones and automotive navigation systems.
In addition to extending the TAM, this research also extends the CFT. This theory
suggests that decision-performance improves when information presentation matches a
problem task (Vessey, 1991). While decision-performance measures were not directly
utilized, this study indicates that hedonic and utilitarian perceptions of a decision-making
133
tool are impacted by the presentation method. In the case of this study, the use of
geovisualization to present geospatial data was examined.
Furthermore, our findings regarding the relationships between the PEOU, PU and
PE on A were consistent with prior research (e.g., Chen et al., 2002; Davis et al., 1992;
Vijayasarathy, 2004).
Finally, this research looks at the impact of A on BI. Consistent with research in
other domains, this research finds that A towards geospatial tools, such as an SDSS, is a
significant predictor of BI to use those tools (e.g., Davis, 1993; Liker and Sindi, 1997;
Mathieson, 1991).
Implications for Industry
As BI has been shown to be a strong indicator of actual use (e.g., Ajzen, 1985;
Ajzen & Fishbein, 1980; Yi, Jackson, Park & Probst, 2006), these results indicate that the
effect of GRA may ultimately influence actual use of systems requiring GRA, such as
SDSS. As many business decisions are made using geospatial data, the ability to
effectively use this type of data for decision-making is essential. This research suggests
that businesses need to understand employee skills in this area and be prepared to provide
support, training and/or more advanced geovisualization tools to help support individuals
with lower GRA. While a demographic understanding of GRA, including percentages of
the population with low or high GRA, as well as cultural, education or gender effects are
yet to be established, the knowledge that GRA impacts technology adoption could
ultimately encourage providers of SDSS technologies to provide visualization
alternatives for specific users.
134
Implications for Scholarship
This research project provided an initial nomological network for the GRA
construct. The establishment of such a network will allow future research to perform
comparative analyses. Furthermore, the TAM was extended to include GRA as an
external variable. While TAM has been applied to numerous technologies and situations,
it had not yet been applied to the conceptual domain of geovisualization.
Prior research has shown that utilitarian measures are a stronger indicator of user-
acceptance of utilitarian systems, while hedonic measures are a stronger indicator of user-
acceptance of hedonic systems (van der Heijden, 2004). The indication that utilitarian and
hedonic measures nearly equally influence attitude toward geovisualization, provides a
better understanding of technologies that encompass both utilitarian and hedonic
attributes. Furthermore, this research expands the knowledge of information systems that
contain combined utilitarian and hedonic aspects.
Limitations
This study has several limitations. First, the GRA construct is a new construct that
has yet to be further explored and exposed to repeated empirical validation. Second, the
concept of geovisualization is broad and subjects may have had differing interpretations
of the meaning of geovisualization. While examples of geovisualization were provided to
subjects, some may have had positive pleasurable experiences with such technologies,
while other subjects may have used geovisualization exclusively in their workplace.
Furthermore, limited empirical research has been conducted on information systems with
both utilitarian and hedonic aspects, so this study functions as an exploratory study.
Finally, the research domain of this study was limited to the use of online mapping
135
services, such as Google Maps. Additional research, specifically addressing more
advanced GIS and SDSS tools, will need to be conducted. Finally, only individual
decision-making, not group decision-making, was explored in this study.
Future Research
Future research will be essential to better understand geospatial reasoning ability
and its impact on technology acceptance and usage. Research is needed to further validate
the geospatial reasoning ability construct and to develop norms for geospatial reasoning
ability with respect to various population groups. In addition to identifying such norms, it
will be essential to determine measurable impacts of geospatial reasoning ability on
outcomes such as decision-performance. The effect of geospatial reasoning ability on an
individual’s ability to perform tasks utilizing geospatial information also needs to be
explored. This analysis of geospatial reasoning ability can include business tasks as well
as other areas that heavily utilize geospatial information, such as logistics, civil aviation,
and military applications. Finally, the effect of geospatial reasoning ability on group
decision-making should be explored.
Conclusion
The TAM, as originally developed by Davis (1989), has been demonstrated to be
an effective method to evaluate the acceptance of various information systems. This
research expands this model to include geospatial reasoning ability and measures
adoption of SDSS.
As both consumers and business decision-makers leverage the benefits of SDSS
to analyze and interpret complex geospatial information, a better understanding of what
factors impact the adoption of SDSS is essential to industry and scholarship. This study
136
provides an exploratory analysis of the user-acceptance of SDSS and identifies several
important findings. First, cognitive ability plays a role in the utilitarian and hedonic
perceptions of SDSS. Second, geospatial reasoning ability can be a predictor of the
behavioral intent to use SDSS, which is an indicator of actual use. Implications of this
research include emphasizing the combined utility of measuring hedonic and utilitarian
perceptions for a technology that contains both hedonic and utilitarian attributes. The
findings of this study inform industry and scholarship of the significance of geospatial
reasoning ability on the hedonic and utilitarian perceptions of information systems, which
ultimately effects adoption of systems designed to improve decision-making using
geospatial data.
137
CHAPTER V
INDIVIDUAL DECISION-PERFORMANCE OF SPATIAL
DECISION SUPPORT SYSTEMS: A GEOSPATIAL REASONING
ABILITY AND PERCEIVED TASK-TECHNOLOGY FIT
PERSPECTIVE4
Abstract
Chapter V continues the validation process of the GRA construct and scale by
incorporating these into an experiment based on the research framework presented in
Chapter II.
Increasingly, spatial decision support systems (SDSS) help consumers, businesses
and governmental entities make decisions involving geospatial data. Understanding if,
and how, user- and task-characteristics impact decision-performance will allow SDSS
developers to maximize decision-making performance. Furthermore, scholars can benefit
from a more comprehensive understanding of what specific characteristics influence
decision-making using an SDSS. CFT is used as the theoretical framework of this study,
along with research in decision-performance and geospatial reasoning ability. This paper
provides a synthesis of research investigating user-characteristics, task-characteristics and
decision-performance when using SDSS. Two hundred subjects participated in a two-
factor experiment designed to measure the impact of user- and task-characteristics on
decision-performance. Specifically, geospatial reasoning ability is investigated as a user-
characteristic, while problem-complexity and presentation-complexity are used as
treatments of task-characteristics. Decision-time and decision-accuracy are used as
4 An early version of this chapter was presented at the Americas Conference on Information Systems
(AMCIS) 2013. A subsequent version is of this chapter is under review at Decision Support Systems.
138
measures of decision-performance, while task-technology fit is applied as an indicator of
user satisfaction. A partial least squares analysis revealed statistical significance of user-
and task-characteristics on decision performance. Theoretical and managerial
implications are discussed in detail.
Introduction
Consumer, government and business decision-makers increasingly rely on SDSS.
For example, consumers often use web-based SDSS for a variety of decision-making
tasks, including locating a nearby bank, selecting the best route to an airport or even more
complex tasks, such as selecting an ideal neighborhood for a new home. In addition,
business decision-makers utilize such tools to assign sales territories, determine sites for
outdoor advertising campaigns and to achieve supply-chain efficiencies. Government and
community groups apply such tools to communicate complicated geospatial concepts, to
determine areas that could be impacted by civic projects, and to spatially correlate social
and natural phenomena with their possible causes. These examples are just a few of the
many ways that entities use SDSS to support geospatial decision-making.
Decision-makers often have access to large quantities of geospatial data, which is
continuously collected using mobile devices and shared from a variety of sources.
Furthermore, several free or low-cost tools allow organizations to easily develop and
provide SDSS technology to decision-makers. As many business decisions utilize
geographic data, understanding how such decisions are made and how such decision-
making can be improved provides an important benefit to organizations (Tonkin, 1992;
Mennecke, 1997).
139
While prior research papers address aspects of decision-performance, user-
characteristics and task-characteristics (Jarupathirun & Zahedi, 2007; Ozimec, Natter, &
Reutterer, 2010), none of these specifically measure geospatial reasoning, problem-
complexity and visualization-complexity with decision-performance simultaneously.
Understanding how user- and task-characteristics influence decision performance when
solving geospatial problems provides an important extension to existing knowledge in
this area. Furthermore, industry will benefit by learning how user- and task-
characteristics can be augmented to enhance decision-performance.
Three primary research questions provide the motivation for this study: (1) Does
geospatial reasoning ability impact geospatial decision-making performance? (2) Does
the complexity of the visualization impact geospatial decision-making performance? (3)
Does the complexity of the problem impact geospatial decision-making performance?
Each of these questions is addressed in this research project.
The following section presents a literature review and theoretical framework for
this study. Subsequently, a comprehensive research model and accompanying research
method are presented. These sections are followed by findings derived through a partial
least squares analysis. Finally, a discussion, limitations, suggestions for future research
and a conclusion are presented.
Literature Review
This research extends the CFT by exploring the effects of user- and task-
characteristics on decision performance. Specifically, the impact of the user-characteristic
of geospatial cognitive ability on decision-performance is examined.
140
Cognitive Fit Theory
Vessey’s (1991) CFT provides the theoretical framework for this study. CFT
suggests that higher quality decisions are made when the information presentation
matches the problem-solving task. The CFT has been highly cited within the information
systems scholarship and has been extended into several decision-making studies
involving geospatial data (e.g., Smelcer & Carmel, 1997; Swink & Speier, 1999; Speier
& Morris, 2003; Mennecke et al., 2000).
Several studies explore the impact of task complexity on decision-making
performance, particularly when examining a problem involving geospatial data (e.g.,
Smelcer & Carmel, 1997; Swink & Speier, 1999; Jarupathirun & Zahedi, 2007; Ozimec
et al., 2010). For instance, Swink and Speier validate that decision-making performance,
as measured by decision-quality and decision-time, is superior for less complex
problems. Additionally, while Mennecke et al. (2000) confirm that as task-complexity
increases, accuracy is lowered, only partial support for task efficiency being lowered was
found. Speier’s (2006) review of research examining cognitive fit, noted that seven of the
eight papers examined provided full or partial support of the CFT.
While there have been extensions of CFT related to geospatial decision-making,
Kelton, Pennington, and Tuttle (2010) state that:
…in order to improve the generalizability of future research, researchers should
attempt to isolate and identify the manner in which problem representations differ
and examine the cognitive effects resulting from these format factors and
interactions among them. (p. 89)
141
Decision Performance
Numerous studies have examined objective decision-performance, measured by
decision-accuracy and decision-time, when making decisions using geographic
information (e.g., Swink & Speier, 1999; Crossland & Wynne, 1994; Crossland et al.,
1995; Smelcer & Carmel, 1997; Dennis & Carte, 1998; Ozimec et al., 2010). For
example, Crossland et al. examined decision-time and decision-accuracy when evaluating
the use of SDSS versus paper maps. Additionally, Dennis and Carte utilized decision-
time and decision-performance when comparing map and tabular geospatial information
presentation. Other decision-performance indicators have included perceptions of the
decision outcome or process, such as perceived decision-quality and perceived decision-
confidence (Jarupathirun & Zahedi, 2007; Ozimec et al., 2010).
Objective measures of decision-making time and decision accuracy are the most
commonly validated measures of decision-making performance. However, research also
suggests that incorporating the use of perceptions of the decision-making process and
performance, particularly as user perceptions have been shown to be significant aspects
of technology acceptance (Jarupathirun & Zahedi, 2007; Ozimec et al., 2010).
Perceived Task-Technology Fit
Additionally, the user satisfaction measure of perceived task-technology fit has
been tested in relation to geospatial decision-performance with significant results
(Jarupathirun & Zahedi, 2007). The perceived task-technology fit measure allows user
satisfaction to be assessed, revealing the potential likelihood of adoption and intention to
use the technology (Rogers, 1983; Taylor & Todd, 1995).
142
User Characteristics
Several studies have investigated individual user-characteristics and their impact
on successfully processing geographic information and making use of it in decision-
making. For example, user characteristics such as sex, age, culture, cognitive ability,
mental workload, spatial visualization, spatial orientation, and general spatial ability have
been examined (Albert & Golledge, 1999; Slocum et al., 2001; Zipf, 2002; Speier &
Morris, 2003; Jarupathirun & Zahedi, 2007). Numerous studies utilize spatial
visualization ability as a measure of user characteristics (Smelcer & Carmel, 1997;
Whitney et al., 2011). Other studies have included spatial orientation (Swink & Speier,
1999), self-efficacy (Jarupathirun & Zahedi, 2007) and concepts including visual memory
and perspective taking (Whitney et al., 2011). Lee and Bednarz (2009) raise a concern
that many cognitive evaluation tools that should evaluate geospatial reasoning are based
on ‘table-top’ measurements and may not actually evaluate cognitive abilities in the
geospatial context. This concern was addressed through the development of a multi-
dimensional construct designed to measure geospatial reasoning ability (Erskine &
Gregg, 2011). Understanding user-characteristics that impact individual decision-making
performance is essential as such knowledge will allow researchers to develop tools and
presentation techniques that improve decision-making performance for those with lower
cognitive ability without impacting those with high cognitive ability (Smelzer & Carmel,
1997).
Lawton (1994) discovered that there were significant differences for spatial
perception, mental rotation, route strategy, orientation strategy and spatial anxiety among
the different gender groups. Specifically, males scored higher on mental rotation, spatial
143
perception and orientation strategy, while females scored higher on route strategy and
spatial anxiety. In their meta-analysis of spatial ability, Linn and Petersen (1985) reported
that mental rotation and spatial perception abilities scores were higher for male subjects.
Additionally, it was reported that gender accounts for up to 5% of performance
differences in most spatial tasks. However, Evans (1980) reported that the impact of
gender had mixed significance impacts based on the task. The author also noted that most
reviewed papers suffered from a limitation in that paper-and-pencil spatial ability tests
were primarily conducted. Yet, Kozlowski and Bryant (1977) reported that subjects who
provided a self-rating of sense of direction had no significant differences due to gender.
Furthermore, spatial ability, as measured through mental rotation, and mediated by
gender has been found to influence multitasking (Mäntylä, 2013). Rusch et al. (2012)
found a significant relationship between gender and analysis accuracy, with male subjects
having a higher accuracy. Furthermore, specific factors such as menstrual cycles and
genetics have been suggested as influencing the impact of gender differences research
(Linn & Petersen, 1985; Mäntylä, 2013).
In addition to user-characteristics, task-characteristics have also been explored in
similar studies.
Task Characteristics
Numerous task characteristics have been evaluated in previous studies, including
map types, map symbolization, (Ozimec et al., 2010) and information presentation
(Dennis & Carte, 1998). However, one of the most common task-characteristic measures
is that of task difficulty (Smelcer & Carmel, 1997; Jarupathirun & Zahedi, 2007; Ozimec
et al., 2010). Task difficulty can be manipulated through the complexity of the
144
relationships of the data analyzed (Smelcer & Carmel, 1997) as well as through the
number of possible solutions and functionality of tools provided (Jarupathirun & Zahedi,
2007).
Table 55 provides an overview of relevant decision-performance studies that
explore task characteristics, user characteristics and decision-performance.
145
Table 55. Summary of Geospatial Decision Performance Research.
Study Task
Characteristics
User
Characteristics
Decision
Performance
Crossland et al.,
1995
Visualization Tool Decision Time,
Decision Accuracy
Smelcer & Carmel,
1997
Task Difficulty,
Geographic
Relationships, Data
Representations
Spatial Visualization Decision Time,
Decision Accuracy
Dennis & Carte,
1998
Information
Presentation
Decision Time,
Decision Accuracy
Swink & Speier,
1999
Problem Size, Data
Aggregation, Data
Dispersion
Spatial Orientation Decision Quality,
Decision Time
Jarupathirun &
Zahedi, 2007
SDSS Functionality,
Site Selection, Task
Complexity, Goal
Difficulty
Visualization,
Spatial Orientation,
Self Efficacy
Perceived Decision
Quality, Perceived
Decision
Efficiency,
Decision
Satisfaction, SDSS
Technology
Satisfaction
Ozimec et al., 2010 Map Type, Map
Symbolization, Task
Complexity, Goal
Difficulty
Spatial Ability, Map
Experience
Decision
Efficiency,
Decision Accuracy,
Decision
Confidence,
Perceived Ease of
Task
Whitney et al., 2011 Address Verification Spatial
Visualization,
Visual Memory,
Perspective Taking
Field Travel
Distance, Total
Time, Number of
Errors
Note: Only measures directly relevant to this study are shown in Table 55; see
original works for further information.
146
Research Model
This research project utilizes geospatial reasoning ability and perceived task-
technology fit as user-characteristics, presentation complexity and problem complexity as
task-characteristics, and decision-time and decision-accuracy as measures of decision-
performance. See Figure 10 for a visual representation of the proposed research model.
This research paper builds upon previous experiments, as shown in Table 55, by
including the multi-dimensional measure of geospatial reasoning ability to measure
spatial cognition within the context of a geographic scale. As demonstrated by the
literature review, objective measures of decision-making are widely-used measures of
decision-making performance.
Geospatial Reasoning Ability(GRA)
Perceived Task-Technology Fit(PTTF)
Decision Accuracy(A)
Decision Time(T)
Problem Complexity(ProbC)
Presentation Complexity(PresC)
H3a
H1a
H2a
H4a
H1c
H3b
H2b
H1b
H4b
Figure 10. Proposed Research Model.
Prior research has shown conflicting results when measuring the effects of spatial
ability on objective and subjective decision-making performance measures. For example,
Albert and Golledge (1999) and Jarupathirun and Zahedi (2007) reported partial or no
147
significant effect of spatial ability on their experiment outcomes. However, Smelcer and
Carmel (1997), Swink and Speier (1999), Speier and Morris (2003), Lee and Bednarz
(2009), Whitney et al. (2011) and Rusch et al. (2012) discovered a significant effect
between spatial ability on the outcomes of their experiments. Many of these studies
utilize measurement instruments that examine cognitive reasoning outside of the
geospatial context, measure only one or two dimensions of spatial reasoning, and often
require some previous SDSS experience. However, the geospatial reasoning ability
(GRA) scale examines such reasoning using three dimensions within the geospatial
context and allows expert as well as non-expert responses (Erskine & Gregg, 2011).
Furthermore, as decision-time and decision-accuracy have been commonly used as
measurements of decision-performance, we suggest measuring the impact of GRA on
these two measures. Based on these findings, we posit:
H1a: Higher geospatial reasoning ability (GRA) leads to lower decision time (T).
H1b: Higher geospatial reasoning ability (GRA) leads to increased decision
accuracy (A).
H1c: Higher geospatial reasoning ability (GRA) leads to higher perceived task-
technology fit (PTTF).
In addition to user characteristics, task characteristics such as task complexity
have been demonstrated to impact decision-making performance. Indeed, one of the most
common measures of task characteristics is that of task difficulty (Smelcer & Carmel,
1997; Jarupathirun & Zahedi, 2007; Ozimec et al., 2010). Several studies cite Campbell’s
(1988) Task Complexity Theory, which suggests that task types have attributes that
influence complexity, as well as suggesting that objective characteristics, psychological
148
experience and task-person interaction also influence complexity (Speier & Morris, 2003;
Jarupathirun & Zahedi, 2007). Finally, support was found linking decision-making
performance and cognitive fit (Vessey, 1991; Jarupathirun & Zahedi, 2007). For instance,
a perceived task-technology fit (PTTF) scale was used to measure individual perceptions
of SDSS performance using a seven-item scale with a statistically significant impact
(Jarupathirun & Zahedi, 2007). Similarly, as decision-time and decision-accuracy have
been commonly used as measurements of decision-performance, we suggest measuring
the impact of PTTF on these two measures. Based on these findings, we posit:
H2a: Higher perceived task-technology fit (PTTF) leads to lower decision time
(T).
H2b: Higher perceived task-technology fit (PTTF) leads to increased decision
accuracy (A).
Finally, research has identified that increases in complexity lead to decreased
decision-making performance (Smelcer & Carmel, 1997). Specifically, increases to the
visualization complexity and task complexity have been shown to increase decision-
making time. Smelcer and Carmel examined three levels of task-difficulty moderated by
the number of sub-tasks each problem required, revealing that increased task complexity
leads to increased decision-time.
Furthermore, Dennis and Carte (1998) tested decision-time and decision-accuracy
on geographic containment and geographic adjacency tasks, revealing decision-
performance improvements when map-based presentations were used (versus tabular
presentations) in all cases, except for decision accuracy when performing geographic
containment tasks. Additionally, Swink and Speier (1999) found that decision-
149
performance, as measured through decision-quality and decision-time, decreased as
problem complexity increased.
Based on these findings, we posit that increasing task complexity leads to
degraded decision-performance, or more specifically:
H3a: Higher problem complexity (ProbC) leads to increased decision time (T).
H3b: Higher problem complexity (ProbC) leads to decreased decision accuracy
(A).
H4a: Higher presentation complexity (PresC) leads to increased decision time (T).
H4b: Higher presentation complexity (PresC) leads to decreased decision
accuracy (A).
Research Methodology
This research project was evaluated using an experiment with a two-by-two
treatment design. Specifically, subjects were asked to perform a geospatial decision-
making task where the problem-complexity and the visualization-complexity were
manipulated. In addition to the experiment, participants were asked to provide
demographic information and complete a geospatial reasoning ability and perceived task-
technology fit measurement scales. Figure 11 demonstrates the experiment workflow as
perceived by the research subjects.
150
Consent Form DemographicsGRA Measurement Scale
(12 Items)
2x2 Experiment Random Assignment
(Decision Time, Decision Accuracy)
PTTF Measurement Scale(7 Items)
Open Feedback
Figure 11. Experiment Workflow.
First, the subject was presented with a consent form. Upon agreement,
demographic information was collected, including age, gender, education and cultural
background. Next, the subject was asked to complete the 12-item GRA measurement
scale. Then the experiment was presented, which collected decision-time and decision-
accuracy data. One of four experiment modes was randomly selected. Presentation
complexity was manipulated by controlling for the number of location items (e.g.,
businesses) that appeared on the map. Problem complexity was controlled based on the
number of decision criteria the subjects were asked to apply to the problem. See Table 56
for a tabular comparison of the experiment modes. Table 60 provides additional detail
regarding each of the modes. Finally, the 7-item PTTF measurement was presented and
the data collection was completed.
Table 56. Summary of Geospatial Decision Performance Research.
Experiment
Mode
Visualization
Complexity Task Complexity
Mode 1 Low/Easy Low/Easy
Mode 2 Low/Easy High/Difficult
Mode 3 High/Difficult Low/Easy
Mode 4 High/Difficult High/Difficult
151
Experiment Design
The experiment asks participants to evaluate a decision-making tool in a
hypothetical scenario in which they must select the ideal apartment for a friend moving to
another country. Detailed evaluation criteria, which include spatial and non-spatial
criteria, are provided (e.g., cost and location preferences). The complexity of the decision
criteria and the problem scale are manipulated to create realistic scenarios, yet still allow
variations in the treatment. Figure 12 presents an overview of the apartment finder tool.
Figure 12. “Apartment Finder” Experiment Tool Developed using GISCloud and
Bing Maps.
Subjects
Responses from 200 subjects were collected from January through May of 2013.
Various methods were employed to solicit subjects, including e-mails and social network
participant recruitment. Participation in the study was voluntary. Some subjects received
a nominal amount of extra credit for participating in the study. The experiment consisted
152
of four experiment modes in which 50 subjects were included for each mode. Descriptive
statistics of the 200-subject pool are presented in Table 57.
Table 57. Descriptive Statistics of Experiment Subjects.
Question Variables Percentage
Age 18-25 48.00%
26-35 37.50%
36-46 10.00%
46-55 3.00%
56-65 1.00%
65+ 0.50%
Gender Female 44.00%
Male 56.00%
Education Elementary/Middle School 0.00%
High School 33.00%
2 Year/Associate Degree 34.00%
4 Year/Bachelor Degree 22.00%
Doctor/JD/PhD 10.00%
Cultural
Background
African 1.00%
Australian 0.00%
Asian 13.00%
European 36.00%
Middle Eastern 20.5%
North American 25.5%
South American 2.00%
Geospatial Reasoning Ability Measurement Items
To assess geospatial reasoning ability the twelve-item geospatial reasoning ability
scale developed by Erskine and Gregg (2011, 2012, 2013) is utilized. This scale
addresses three-dimensions of geospatial reasoning ability: (1) geospatial memorization
and recall, (2) geospatial orientation and navigation and (3) geospatial visualization.
153
These measurement items, as shown in Table 58, are presented prior to the experiment
along with a seven-item Likert scale (Likert, 1932).
Table 58. Geospatial Reasoning Ability Measurement Items, Adapted from Erskine
and Gregg (2011, 2012, 2013).
Item ID Item
SPGMR1 I can usually remember a new route after I have traveled it only
once.
SPGMR2 I am good at giving driving directions from memory.
SPGMR3 After studying a map, I can often follow the route without needing
to look back at the map.
SPGMR4 I am good at giving walking directions from memory.
SPGON3 In most circumstances, I feel that I could quickly determine where I
am based on my surroundings.
SPGON4 I have a great sense of direction.
SPGON5 I feel that I can easily orientate myself in a new place.
SGON10 I rarely get lost.
SPGV1 I can visualize geographic locations.
SPGV2 I can visualize a place from information that is provided by a map
without having been there.
SPGV3 I can visualize a place from a map.
SPGV10 While reading written walking directions, I often form a mental
image of the walk.
Perceived Task-Technology Fit Measurement Items
In addition to decision-time and decision-accuracy, perceived task-technology fit
was measured as an indicator of user satisfaction toward decision-performance. To assess
perceived task-technology fit, measurement items developed by Karimi et al. (2004) and
Jarupathirun and Zahedi (2007) were adapted for this study. These items were presented
to the research subjects upon completion of the decision-making experiment along with a
seven-point Likert scale. Table 59 presents the seven-item perceived task-technology fit
measurement scale used for this study.
154
Table 59. Perceived Task-Technology Fit Measurement Items, Adapted from
Karimi et al. (2004) and Jarupathirun and Zahedi (2007).
Item ID Item
PTTF1 The functionalities of the [tool] were adequate for the task given.
PTTF2 The functionalities of the [tool] were appropriate for the task given.
PTTF3 The functionalities of the [tool] were useful for the task given.
PTTF4 The functionalities of the [tool] were compatible with the task
given.
PTTF5 The functionalities of the [tool] were helpful in solving the task
given.
PTTF6 The functionalities of the [tool] were sufficient.
PTTF7 The functionalities of the [tool] made the task easy.
Problem Complexity
Problem complexity was manipulated for each treatment to include either a low or
high complexity mode. For the modes using a low problem complexity, the apartment
finder experiment provided three proximity criteria and one attribute criterion for the
decision task. Alternately, for the modes using a high problem complexity, the apartment
finder experiment required a decision to be made using four proximity criteria and two
attribute criteria.
In the experiment, subjects were asked to find an apartment for a friend in the
small town of Bad Tölz, Germany using the online apartment finder tool. To guide the
decision-making process, a series of spatial- and attribute-based apartment criteria were
provided. The spatial-based criteria consisted of statements regarding visual proximity.
An example of a spatial-based criterion used in both the high and low problem
complexity treatments is: “there must be a grocery store nearby.” To determine
proximities between the map elements, subjects had to estimate the proximity visually.
155
The attribute-based criteria required subjects to select an individual point on the map and
then review its associated attribute table. An example of an attribute criterion, as used in
the high complexity treatment, was: “the apartment rent must be less than 300
Euros/month.”
Table 60. Problem Complexity
Problem
Specification
Assume that a friend has asked you to find an apartment in Bad
Tölz, Germany. The city is testing an online map tool, which should
help you find the best apartment for your friend. Please find the best
apartment based on the following criteria:
Low
Problem
Complexity
Criteria
1) The location should be as close to the Business School Building
(American International University) as possible
2) There must be a grocery store nearby
3) There must be a laundry nearby
4) Additionally, the apartment rent must be less than 350
Euros/month
High
Problem
Complexity
Criteria
1) The location should be as close to the Business School Building
(American International University) as possible
2) There must be a grocery store nearby
3) There must be a laundry nearby
4) The laundry must offer coin-laundry and dry cleaning services
5) Additionally, the apartment rent must be less than 300
Euros/month
6) An apartment with a bar, or other nightlife, nearby would be
preferred
Visualization Complexity
In addition to problem complexity, visualization complexity was also manipulated
to include a low or high complexity treatment. Specifically, the visualization complexity
modified the number of points, or nodes, represented on the map. The low visualization
complexity treatment consisted of forty-two nodes (comprising twelve apartments, six
grocery stores, six pubs, six coin laundries, six university buildings and six restaurants).
The high visualization complexity treatment consisted of eighty-four nodes (comprising
156
twenty-four apartments, twelve grocery stores, twelve pubs, twelve coin laundries, twelve
university buildings and twelve restaurants).
Analysis
To perform the statistical analysis, a structural equation modeling/partial lease
squares (SEM-PLS) analysis was conducted using SmartPLS (Ringle et al., 2005). Both
the measurement and structural models of the proposed research model were evaluated.
The PLS algorithm was set to use a path weighting scheme and a maximum of 300
iterations. Upon successfully completing the algorithm, an evaluation of the stop criterion
changes revealed that the algorithm converged after Iteration 3 of the SmartPLS analysis.
To perform the analysis, both the measurement model and structural model were
assessed. For the measurement model assessment, the composite and indicator reliability
as well as the convergent and discriminant validity were assessed.
Measurement Model
The first step of the measurement model evaluation involved testing the construct
reliability using Cronbach’s (1951) alpha and composite reliability. It is recommended
that both of these measures have values above 0.70, as higher values indicated reliability
(Cronbach, 1951; Gliem & Gliem, 2003). However, it is also noted that values above
0.95 can indicate that measurement items are too similar. Previous studies have
demonstrated that the various items comprising the GRA construct are designed to
measure three distinct substrata of GRA, including geospatial memorization and recall,
geospatial orientation and navigation, as well as geospatial visualization (Erskine &
Gregg, 2011, 2012). Furthermore, while the PTTF measurement items are quite similar,
they have been successfully used in previous studies (Karimi et al., 2004; Jarupathirun &
157
Zahedi, 2007). The Cronbach’s alpha and composite reliability of the GRA and PTTF
constructs are shown in Table 61.
Table 61. Cronbach’s alpha and Composite Reliability.
Construct Cronbach’s alpha Composite Reliability
GRA 0.9463 0.9534
PTTF 0.9782 0.9817
To further evaluate the measurement model, both indicator reliability and
convergent validity were assessed. To determine indicator reliability, the loading of each
measurement item on its respective construct was evaluated. Generally, it is suggested
that measurement items have a loading above 0.70, which all items of the GRA construct,
except GRA12, demonstrated. However, due to evidence of this item’s value toward the
GRA construct’s content validity, it was retained. Additionally, all measurement items of
PTTF were above this threshold. The measurement item loadings on each construct are
shown in Table 62.
158
Table 62. Measurement Item Loadings.
Construct GRA PTTF
GRA1 0.7458
GRA2 0.8269
GRA3 0.8101
GRA4 0.8025
GRA5 0.8153
GRA6 0.8959
GRA7 0.8176
GRA8 0.8095
GRA9 0.8000
GRA10 0.8371
GRA11 0.8010
GRA12 0.5333
PTTF1 0.9484
PTTF2 0.9476
PTTF3 0.9497
PTTF4 0.9451
PTTF5 0.9431
PTTF6 0.9290
PTTF7 0.9205
Following the establishment of indicator reliability, convergent validity was tested
by evaluating the average variance extracted (AVE). For this test, the AVE of each
construct should exceed the 0.50 threshold (Fornell & Larcker, 1981). The AVE for GRA
and PTTF were 0.6332 and 0.8846, respectively, indicating convergent validity. See
Table 63 for the AVE of each multi-item, reflective construct.
159
Table 63. Average Variance Extracted (AVE) by Construct.
Construct AVE
GRA1 0.6332
PTTF 0.8846
T N/A – Formative, Single Item
A N/A – Formative, Single Item
ProbC N/A – Formative, Single Item
PresC N/A – Formative, Single Item
Finally, the measurement model was tested for discriminant validity using two
methods. First, the cross-loadings of measurement items were compared between the
GRA and PTTF construct. Generally, measurement items of a construct should have a
loading of above 0.50 (Hair et al., 2010), while above 0.70 would be ideal (Hair, Hult,
Ringle, & Sarstedt, 2014). Using this test, all items indicate convergent validity. See
Table 64 for measurement item cross-loadings.
160
Table 64. Measurement Item Cross-Loadings.
Construct GRA PTTF
GRA1 0.7458 0.3516
GRA2 0.8269 0.3299
GRA3 0.8101 0.3479
GRA4 0.8025 0.3040
GRA5 0.8153 0.4032
GRA6 0.8959 0.3882
GRA7 0.8176 0.3326
GRA8 0.8095 0.2947
GRA9 0.8000 0.3219
GRA10 0.8371 0.2560
GRA11 0.8010 0.2495
GRA12 0.5333 0.0697
PTTF1 0.3627 0.9484
PTTF2 0.4152 0.9476
PTTF3 0.3695 0.9497
PTTF4 0.3700 0.9451
PTTF5 0.3493 0.9431
PTTF6 0.3791 0.9290
PTTF7 0.3692 0.9205
Highest loadings shown in bold.
A second test of discriminant validity compares the square root of each
construct’s AVE with the highest correlation among the other constructs. The square root
of the AVE of GRA was 0.7957 and of PTTF was 0.9405. As these values are larger than
any of the construct correlations, discriminant validity was further indicated. See Table
65 for the complete results of discriminant validity testing.
161
Table 65. Latent Variable Correlations and Sq. of AVE (shown in bold).
A GRA PTTF PresC ProbC T
A N/A
GRA 0.5086 0.7957
PTTF 0.5265 0.3974 0.9405
PresC -0.1908 0.0176 -0.0240 N/A
ProbC -0.1305 -0.0720 0.0343 0.0000 N/A
T -0.4803 -0.5579 -0.5702 0.1069 0.1555 N/A
Structural Model
Following the measurement model evaluation, the structural model was evaluated.
The first step of this evaluation was the examination of collinearity using each predictor
construct’s tolerance and variance inflation factor (VIF). The VIF, or the reciprocal of
tolerance, is defined mathematically as 1/(1-R2). Scholars have recommended maximum
VIF values of 10, which each construct’s VIF meets. Thus, based on the statistical
analysis of VIF, the structural model does not exhibit collinearity. See Table 66 for the
R2, tolerance and VIF values of the three endogenous latent variables (Hair et al., 2014).
See Figure 13 for the visual results of the SmartPLS SEM-PLS algorithm.
Table 66. R2 Values of Endogenous Latent Variables.
Construct R2 Tolerance 1-R
2 VIF (1/(1-R
2))
PTTF 0.158 0.842 1.188
A 0.433 0.567 1.764
T 0.486 0.514 1.946
162
Figure 13. Visual result of SEM-PLS Algorithm (using SmartPLS).
Next, the significance of each path coefficient was evaluated using the
bootstrapping method. For this analysis, the number of cases was set to 200, which are
identical to the total number of valid observations, and the number of samples was set to
5000 as recommended by Hair et al. (2014). All hypothesis tests fell above Gefen and
Straub’s (2005) recommended t-statistic threshold of 1.96. See Table 67 for complete
significance results of the path coefficients.
Table 67. Path Coefficients and Significance Levels.
Hypotheses Path Path
Coefficients
T Values Significance
Levels
H1a (-) GRA T -0.3823 6.7321 p < .01
H1b (+) GRA A 0.3495 6.9855 p < .01
H1c (+) GRA PTTF 0.3974 6.2971 p < .01
H2a (-) PTTF T -0.4207 6.5972 p < .01
H2b (+) PTTF A 0.3871 6.8840 p < .01
H3a (+) ProbC T 0.1424 2.8014 p < .01
H3b (-) ProbC A -0.1187 2.2055 p < .05
H4a (+) PresC T 0.1035 1.9859 p < .05
H4b (-) PresC A -0.1876 3.4969 p < .01
163
Next, significance testing of the total effects is shown. Here again, all paths show
significance levels of 0.01, except H3b and H4a, which had a statistical significance of
0.05. See Table 68 for complete significance results of the total effects, and Figure 14 for
the visual results of the SmartPLS bootstrapping algorithm.
Table 68. Path Coefficients and Significance Levels.
Hypotheses Path Path
Coefficients
T Values Significance
Levels
H1a (-) GRA T -0.5495 10.2458 p < .01
H1b (+) GRA A 0.5034 10.6981 p < .01
H1c (+) GRA PTTF 0.3974 6.2971 p < .01
H2a (-) PTTF T -0.4207 6.5972 p < .01
H2b (+) PTTF A 0.3871 6.8840 p < .01
H3a (+) ProbC T 0.1424 2.8014 p < .01
H3b (-) ProbC A -0.1187 2.2055 p < .05
H4a (+) PresC T 0.1035 1.9859 p < .05
H4b (-) PresC A -0.1876 3.4969 p < .01
Figure 14. Visual result of SEM-PLS Algorithm (using SmartPLS).
164
The next examination involved an analysis of the Coefficients of Determination
or R2 values. GRA, PTTF, ProbC and PresC jointly explain 48.6% of the variance of T
and 43.3% of the variance of A. These values can be considered close to moderate.
Finally, GRA alone explains 15.8% of the variance of PTTF, which is weak. Following
this, the blindfolding procedure of SEM-PLS was used to determine predictive relevance
or Q2 of the endogenous constructs. The Q
2 values revealed large predictive relevance of
A and T, yet only an approximately medium relevance of PTTF. The R2 and Q
2 values of
the endogenous latent variables are shown in Table 69.
Table 69. R2 and Q
2 Values of Endogenous Latent Variables.
Latent Variable R2 Q
2
PTTF 0.158 0.1387
A 0.433 0.4321
T 0.486 0.4749
Heterogeneity
The final test analyzes the data set for heterogeneity, which is of concern as
gender has previously been shown to impact spatial reasoning. First, statistical
differences between male (n=112) and female (n=88) participants were estimated. To do
so, responses were grouped by male and female population selections. The path
coefficients and t-values of the population selection consisting of only male subjects are
shown in Table 70.
165
Table 70. Path Coefficients – Male Only.
Hypotheses Path Path
Coefficients
t-Values Standard Error
H1a (-) GRA T -3.1076 9.4045 0.0588
H1b (+) GRA A 1.2333 7.0510 0.0655
H1c (+) GRA PTTF 0.4078 3.4545 0.0765
H2a (-) PTTF T -1.4922 3.8256 0.0896
H2b (+) PTTF A 0.6853 4.1034 0.0788
H3a (+) ProbC T 0.2035 1.4306 0.0712
H3b (-) ProbC A -0.1355 1.8151 0.0767
H4a (+) PresC T 0.2279 1.6553 0.0687
H4b (-) PresC A -0.1355 2.0429 0.0749
Next, female participant values were retrieved. The path coefficients and t-values
of the population selection consisting of only female subjects are shown in Table 71.
Table 71. Path Coefficients – Female Only.
Hypotheses Path Path
Coefficients
t-Values Standard Error
H1a (-) GRA T -1.5434 2.5473 0.1054
H1b (+) GRA A 0.8158 3.7583 0.0780
H1c (+) GRA PTTF 0.7141 5.2503 0.0892
H2a (-) PTTF T -2.0264 5.4614 0.0984
H2b (+) PTTF A 0.8544 5.7269 0.0817
H3a (+) ProbC T 0.3967 2.7068 0.0712
H3b (-) ProbC A -0.0926 1.2248 0.0759
H4a (+) PresC T 0.1501 0.9448 0.0767
H4b (-) PresC A -0.2229 2.8430 0.0782
Finally, the groups are compared. A test for equality of standard errors between
the male and female groups revealed significant (alpha = 0.10) results for all but one
hypothesis, H1a. See Table 72 for the complete results.
166
Table 72. Result of Gender Group Comparison.
Hypotheses Path Equal standard errors
assumed.
Unequal standard
errors assumed.
Test for
equality
of
standard
errors
t-
value
df p-
value
t-
value
df p-
value
H1a (-) GRA
T
13.728 198 0.000 13.031 137 0.000 1.000
H1b (+) GRA
A
4.147 198 0.000 4.120 180 0.000 0.702 ^
H1c (+) GRA
PTTF
2.630 198 0.009 2.620 182 0.010 0.629 ^
H2a (-) PTTF
T
4.021 198 0.000 4.035 187 0.000 0.394 ^
H2b (+) PTTF
A
1.482 198 0.140 1.497 191 0.136 0.203 ^
H3a (+) ProbC
T
1.901 198 0.059 1.929 193 0.055 0.118 ^
H3b (-) ProbC
A
0.393 198 0.695 0.400 194 0.690 0.099 ^
H4a (+) PresC
T
0.758 198 0.449 0.759 186 0.449 0.458 ^
H4b (-) PresC
A
0.804 198 0.423 0.811 191 0.418 0.222 ^
(^ alpha = 0.10).
Findings and Discussion
The analysis revealed several key findings. The positive user-characteristic of
geospatial reasoning ability impacts decision-performance positively, as measured using
decision-time and decision-accuracy. Furthermore, the task-characteristics of problem
complexity and presentation complexity impact decision-performance, as measured using
decision-time and decision-accuracy. Additionally, a statistically significant relationship
between geospatial reasoning ability and perceived task-technology fit exists, however
167
the magnitude is negligible. See Figure 15 for a visual representation of the significant
relationships within the research model.
Geospatial Reasoning Ability(GRA)
Perceived Task-Technology Fit(PTTF)
Decision Accuracy(A)
Decision Time(T)
Problem Complexity(ProbC)
Presentation Complexity(PresC)
H3a *** (+)
H1a *** (-)
H2a *** (-)
H4a *
* (+
)
H1c *** (+)
H3b *** (-)
H2b *** (+)
H1b *** (+)
H4b *** (-)
**p<0.05, ***p<0.01
Figure 15. Research Model with Relationship Significance.
The hypotheses are well supported as shown in Table 73. These findings suggest
that geospatial reasoning ability and perceived task-technology fit do indeed have an
effect on decision-performance. Thus, these two measures could provide value to those
evaluating SDSS technologies. The impact of geospatial reasoning ability on decision
performance aligns with the results of Smelcer and Carmel (1997), Swink and Speier
(1999), Speier and Morris (2003), Lee and Bednarz (2009), Whitney et al. (2011) and
Rusch et al. (2012), who all discovered a significant effect of spatial ability on the
outcomes of their experiments. Also, the significant effects between perceived task-
technology fit align with the findings of Jarvenpaa (1989) and Vessey (1991) who found
support between decision-making performance and task-technology fit.
168
Table 73. Hypotheses Test.
Hypotheses Path Findings
H1a (-) GRA T Significant/Negative***
H1b (+) GRA A Significant/Positive***
H1c (+) GRA PTTF Significant/Positive***
H2a (-) PTTF T Significant/Negative***
H2b (+) PTTF A Significant/Positive***
H3a (+) ProbC T Significant/Positive***
H3b (-) ProbC A Significant/Negative**
H4a (+) PresC T Significant/Positive**
H4b (-) PresC A Significant/Negative***
**p<0.05, ***p<0.01
In addition, the results support that there are significant differences in the
relationship between geospatial reasoning ability and task complexity for subjects of
different gender for all areas except decision time. For the most part, the significant
differences between gender groups align with previous findings (Linn & Petersen, 1985;
Lawton, 1994). Various reasons could explain why no significance was found for H1a
between gender groups (Linn & Petersen, 1985; Mäntylä, 2013). For instance, the many
experiment and subject factors, such as the spatial tests used in previous studies, which
were often one-dimensional as opposed to the three-dimensional GRA construct, could
account for the lack of significant differences between GRA and decision-time based on
gender groups.
Limitations
This study includes several limitations. For instance, the experiment design
specifically addresses a consumer-oriented decision problem that may be too simple for
generalization to large-scale business and governmental decision-making using SDSS.
Furthermore, problem complexity is manipulated only through the number of proximity
169
relationship and attributes required for each problem. There are many additional methods
for manipulation complexity in geospatial problem solving, such as by including tasks to
determine if points or areas equal, contain, overlap or are located within other areas. The
DE9-IM model developed by Clementini, Di Felice, and van Oosterom (1993) provides
an overview of the various geospatial relationships that exist. Future extensions of this
study should address these geospatial relationships in addition to the proximity
relationships used in this study. Furthermore, only point data was included in the visual
presentation, whereas line and polygon data could have provided a richer problem
solving experiment. Additionally, this study only addresses individual perceptions related
to SDSS decision-making performance. While this is suitable for most consumer
geospatial decision-making, many business and governmental geospatial decision-making
may require group participation.
While this study provides an initial assessment of the impact of geospatial
reasoning ability on decision-making performance, future studies will need to be
conducted in order to better understand geospatial reasoning ability, geospatial decision-
making and their interaction. Additionally, subjects included in this study were primarily
students at an American research institution, so specific cultural and regional aspects of
decision-making may not have been captured. Furthermore, while geospatial reasoning
ability and perceived task-technology fit play a role in the decision-making process, there
are numerous other factors that may influence the process, which have not been
addressed.
170
Implications
Scholarly researchers can benefit from this study for three key reasons. First, the
decision sciences research domain benefits from the additional perspectives regarding
decision-making using SDSS. Second, no prior research utilizing both visualization
complexity and problem complexity as measures of task-complexity were found, thus a
research gap is addressed. Furthermore, a better understanding of the impact of problem-
and presentation- complexity of geovisualized information extends the CFT. Third, the
GRA measurement scale is further empirically validated, making it a viable alternative to
other spatial tests that may not measure multiple dimensions, may not provide a
geographic context or that require previous SDSS experience. These benefits could
provide tremendous benefits to future decision-performance research in the context of
SDSS. Furthermore, such knowledge will provide a foundation for information systems
researchers to develop tools and techniques that improve decision-performance of
individuals with low geospatial reasoning ability.
Industry can greatly benefit from this study as it highlights the importance of
considering geospatial reasoning ability when developing and designing SDSS tools.
Such knowledge could allow managers to allocate individuals with high geospatial
reasoning ability to tasks involving problem-solving using SDSS. Furthermore, a more
comprehensive understanding of the benefits and potential drawbacks of task complexity
(moderated through visualization and problem complexity, herein) can guide systems
designers and developers to help select the most appropriate visualization method for
visualizing complex geospatial relationships.
171
Conclusion
Consumer, business and governmental entities increasingly rely on SDSS for
decision-making involving geospatial data. Understanding user- and task-characteristics
that impact decision performance will allow developers of such systems to maximize
decision-making performance. While there have been several studies that explored the
impact of task- and user-characteristics on decision performance, there have been
inconsistent results. This study explores the potential reasons for these inconsistencies
and provides a comprehensive design to eliminate such problems in future research. For
instance, a two-factor experiment design was implemented to determine the role of
problem-complexity and presentation-complexity on decision-performance. We feel that
a stronger understanding of the characteristics that influence decision-performance when
using SDSS can guide future research in the decision sciences domain.
Finally, this paper extends the nomological network of the geospatial reasoning
ability construct, which has not yet been applied to decision-performance. A further
understanding of the statistically significant relationships of this construct will allow it to
be applied in future studies with greater confidence in its ability to measure geospatial
reasoning ability.
172
CHAPTER VI
CONCLUSION: TOWARD A COMPREHENSIVE
UNDERSTANDING OF GEOSPATIAL REASONING ABILITY AND
THE GEOSPATIAL DECISION-MAKING FRAMEWORK
Abstract
As consumer, business and governmental entities increasingly utilize geospatial
data when making procedural, organizational and strategic decisions, it is essential to
better understand how such decisions are made. This chapter summarizes research
involving the conceptual decision-making framework, in addition to exploring the
nomological network of geospatial reasoning ability (GRA). This summary presents
benefits to industry and academic research. For instance, understanding user- and task-
characteristics that impact decision performance allows developers spatial decision
support systems (SDSS) to maximize decision-making performance. Additionally,
information systems scholars will benefit from a comprehensive understanding of what
specific characteristics influence decision-making, in this case geospatial decision-
making.
Introduction
The previous five chapters provided the initial steps to develop a more
comprehensive understanding of geospatial decision-making within the information
systems scholarship. Chapter I provided a background and motivation for this
dissertation. Chapter II provided a complete literature review and research framework for
geospatial decision-making research. Chapter III suggested the development of a
comprehensive construct defining individual GRA, a key user-characteristic that has
provided mixed results in previous empirical studies. Chapter IV presented an extension
173
of the Technology Acceptance Model in the context of geospatial visualization and the
use of online mapping services. Chapter V explored decision-making within the research
framework presented in Chapter II. This chapter presents a conclusion to the overall
empirical testing of the research framework presented in Chapter II as well as presenting
a nomological network of the GRA construct presented in Chapter III.
While a large body of literature explores geographic visualization, only limited
research explores how such visualizations impact decision-performance, and within
those, conflicting results are often presented. Thus, a simple, yet comprehensive model
for the study of the effects of information presentation on decision-performance was
developed. The core elements of this conceptual model are discussed next.
Conceptual Model
To better understand how decision-making using geospatial data occurs, a broad
conceptual model was developed in Chapter III (see Figure 16). This model was derived
from existing theory, particularly CFT (Vessey, 1991). While the model is initially
applied to geospatial decision-making, it could also be applied to any decision-making
task. The model consists of three propositions:
Proposition P1: Information presentation impacts decision-performance.
Proposition P2: Task-characteristics impact decision-performance.
Proposition P3: User-characteristics impact decision-performance.
174
Figure 16. Conceptual Model.
The model suggests that when task-characteristics, user-characteristics and
information-presentation align, decision-performance is maximized. As this model was
initially applied to geospatial decision-making, measures that have previously been
shown to impact geospatial decision-making were used. In the case of this dissertation,
geospatial visualization using interactive thematic maps is used for information
presentation. Multi-criteria tasks, including proximity information, are used for task-
characteristics. GRA is used as a representation of user-characteristics. Finally, objective
measures of decision-time and decision-accuracy, as well as perceived task-technology
fit, are used to assess decision performance. While these measures are appropriate for
geospatial decision-making, other measures should be applied, as appropriate, based on
relevant theory and literature.
175
Implication to Research
This dissertation presents several important implications to research. First, it
provided a literature review of research concerning decision-making using geospatial
data. This literature review can be referenced by scholars interested in exploring this
important area, and highlights limitations of past studies in addition to several potential
research ideas. Second, CFT was extended to develop a comprehensive decision-making
using geospatial data research model. This model will allow future research regarding
decision-making using geospatial data to apply an appropriate theoretical frame. Third, a
new construct and its associated measurement scale were introduced. This scale was
designed to address limitations of previous scales designed to measure geospatial
reasoning. Furthermore, this scale was subjected to comprehensive statistical testing to
ensure reliability and validity. Fourth, the TAM was extended to include GRA as an
antecedent, demonstrating that cognitive user-characteristics influence technology
adoption. Fifth, the hedonic and utilitarian nature of geovisualization was discovered,
proving future research another applied domain to further explore technologies that
incorporate both of these aspects. Sixth, the GRA construct and the conceptual model of
geospatial decision-making were tested empirically using an experiment. This study
revealed important insights, including the impact of GRA on decision-performance and
provided the first test of the conceptual model developed by extending the CFT.
Implication to Industry
This dissertation also presents numerous benefits to industry. For instance, the
cognitive ability of GRA has demonstrated itself to be important to geovisualization
technology adoption as well as to decision-making using geospatial data. As
176
geovisualization becomes a component of everyday technology interaction, such as
locating a restaurant or a business contact’s address, developing techniques to mitigate
low GRA will be essential for application developers. In addition to an understanding of
GRA, this dissertation also revealed several concepts that will be essential to developers
of location-based services. For instance, significant differences in the problem solving
based on gender could influence marketing and development of such tools. Also, the
specific application of an apartment finding exercise reveals specific implications to
providers of such services, but most importantly that even with multiple criteria,
individuals were able to quickly locate ideal apartments within minutes. Prior to recent
geovisualization technology innovation, such a task would have taken hours if not days.
Additionally, as numerous industries perform such decision-making, these
findings are of great value. For instance, organizational leaders can assemble teams that
have members with greater GRA through the administration of the GRA scale during the
team member selection process. Furthermore, human resources departments could assess
the GRA of individuals applying for positions that require a great deal of individual
spatial ability, such as in logistics, maritime and aviation.
Theoretical Frameworks
In the course of this dissertation, the conceptual model was linked to three
theoretical frameworks: CFT, Task Complexity Theory and the Technology Acceptance
Model.
First, Vessey’s (1991) CFT provided the initial theoretical framework for the
model. This theory suggests that higher quality decisions are made when the information-
presentation matches the problem-solving tasks. CFT has been referenced as a theoretical
177
background, extended into other domains and validated in numerous empirical studies
involving geospatial data, including Smelcer and Carmel (1997), Mennecke et al. (2000),
and Speier and Morris (2003). For instance, Mennecke et al. extended the CFT in an
exploration of how user-characteristics and task-complexity impact decision-
performance.
Second, Campbell’s (1988) Task Complexity Theory is applied when specifically
addressing the task-characteristic relationship within the conceptual model. Task
Complexity Theory suggests that each type of task has attributes that influence
complexity and that complexity is further influenced by user-characteristics, such as
psychological experience (Larson & Czerwinski, 1998). Swink and Speier (1999) applied
Task Complexity Theory to explain that different tasks types may be more or less
complex, ultimately influencing decision-performance.
Finally, Davis’ (1989) Technology Acceptance Model (TAM) is applied when
addressing the user-characteristic relationship within the conceptual model. This model,
which is an adaption of the Theory of Reasoned Action (Fishbein & Ajzen, 1975), is
widely used in the information systems scholarship with the intent to explain why
individuals adopt and use technologies. While the TAM has been revised several times,
the original model along with the hedonic measure of perceived enjoyment was applied
to the use of online mapping services. Specifically, the TAM was applied to determine if
GRA influences the acceptance of SDSS, as well as if SDSS users adopt such systems
based on hedonic, utilitarian properties or a combination of both.
178
Findings and Discussion
Following the brief introduction presented in Chapter I, Chapter II presented a
detailed introduction to the conceptual model described above. Specifically, an argument
was made for use of the model when exploring decision-making using geospatial data.
The literature review revealed that existing research in information systems, as well as
other domains, provides a strong foundation toward the exploration of decision-making
using geospatial data. In addition to the literature review, a research framework was
presented along with a conceptual research model. In addition to the conceptual research
model, a discussion of research questions, theoretical lenses, existing constructs, potential
limitations, and validity concerns appropriate to future research were presented. It is
suggested that the framework could be used as a benchmark for future research, provide a
stronger theoretical background, and facilitate improved comparison and understanding
between studies. This would be tremendously beneficial due to the numerous
contradictions in past research, particularly surrounding task complexity and spatial
reasoning.
Additionally, while business school curriculums include basic computing courses
and courses that address databases and advanced spreadsheets, many of these courses do
not include geospatial problem-solving exercises. Thus, another key proposal of Chapter
II was the suggestion that business school curriculums offer, at a minimum, problem-
solving exercises utilizing geospatial data. Pick (2004) learned that some business
schools included a study of GIS as either an elective or required course; however, with
the prevalence of geospatial data and its increased use in business problem-solving,
perhaps this is not enough.
179
Due to prior research that suggests spatial ability plays an important role in
decision-making ability when working with geospatial data, a new construct measuring
GRA was introduced in Chapter III. While research into the role of cognitive abilities has
been extensive, there have been conflicting results. It is suggested that the discrepancies
are related to the specific dimensions of spatial ability assessed, the lack of geospatial
context, and the need for previous geospatial or tool experience. As IS research has
generally only measured one or two dimensions of spatial ability in each study, the new
GRA construct assesses a spectrum of spatial abilities. The scale development procedures
provided by MacKenzie et al. (2011), were utilized to develop a multi-dimensional scale
with demonstrated reliability and validity.
Furthermore, Chapter III revealed an independent construct that is different from
GRA, but one that may also have an impact on decision-performance. This construct,
called self-perceived geospatial schematization, is developed and presented along with
several suggestions for future research.
In Chapter IV, the newly established GRA construct was further validated. For
this validation step, the Technology Acceptance Model (TAM) was selected as a
theoretical lens, as it is one of the most widely cited information systems models. For
instance, as of August 4, 2013, Google Scholar lists 16,760 citations of the original Davis
(1989) paper alone. The initial GRA construct developed in Chapter III was tested along
with an existing construct measuring perceived enjoyment (PE). Perceived enjoyment
was included due to previous research suggesting that certain technologies contain both
utilitarian and hedonistic attributes that could affect technology adoption. While TAM
has been demonstrated to be an effective method to evaluate the acceptance of various
180
information systems, it had not been applied to SDSS or GIS, nor has GRA been used as
an antecedent. Thus, Chapter III extended TAM in two important ways: 1) GRA was
used as an antecedent of utilitarian and hedonistic perception, 2) TAM was applied in the
context of SDSS. This research chapter revealed two key findings. First, cognitive ability
plays a role in the utilitarian and hedonic perceptions of SDSS. Second, GRA can be a
predictor of the behavioral intent to use SDSS, which is an indicator of actual use. These
findings demonstrate that users of SDSS may perceive such tools as having both hedonic
and utilitarian attributes. Scholarly implications of this research include a confirmation of
the value of measuring the combined hedonic and utilitarian perceptions when assessing a
technology that contains both hedonic and utilitarian attributes. The findings of this study
inform industry and scholarship of the significance of GRA on the hedonic and utilitarian
perceptions of information systems, which ultimately effects adoption of systems
designed to improve decision-making using geospatial data.
In Chapter V, the conceptual model proposed in Chapter II was applied to a
geospatial decision-making experiment. This experiment explored the potential reasons
for inconsistencies in studies that explored the impact of task-characteristics and user-
characteristics on decision performance. Furthermore, great care was taken in the study
design to eliminate external influences. For instance, a two-factor experiment design was
implemented to determine the moderating role of problem-complexity and presentation-
complexity on decision-performance. A stronger understanding of the characteristics that
influence decision-performance of SDSS can guide future research in the decision
sciences domain. Understanding user- and task-characteristics that impact decision-
181
performance will allow developers of such systems to maximize decision-making
performance.
Two hundred subjects participated in a two-factor experiment designed to
measure the impact of user- and task-characteristics on decision-performance.
Specifically, GRA and perceived task-technology fit were investigated as user-
characteristics, while problem-complexity and map-complexity were used as treatments
of task-characteristics. Decision-time and decision-accuracy were used as measures of
decision-performance. A partial least squares analysis revealed statistical significance of
user- and task-characteristics on decision performance.
Chapter V also extended the nomological network of the GRA measurement
scale, which had not yet been applied to decision-performance. A complete nomological
network presenting the empirical results of Chapter IV and V are presented in Figure II.
A further understanding of the statistically significant relationships of this construct’s
measurement scale will allow it to be applied in future studies with greater confidence in
its ability to truly detect GRA. This nomological network should be expanded to include
antecedents of GRA in order to allow tests of predictive efficiency and mediating
efficiency to be performed (Liu, Li, & Zhu, 2012).
182
Geospatial Reasoning Ability(GRA)
Perceived Ease-of-Use
Decision Accuracy
Decision TimePerceived Enjoyment
Perceived Usefulness
Perceived Task-Technology Fit
Figure 17. Current Nomological Network of GRA Construct.
Future Research
Finally, this chapter concludes with a discussion of the limitations and
recommendations for future research related to both the GRA construct as well as the
conceptual decision-making research model.
Future scholarly research should include further validation and establishment of
norms for the GRA construct. It is suggested that GRA measurements be compared with
well-known spatial reasoning tests, such as VZ1. Such future research will be essential to
better understand geospatial reasoning ability and its impact on technology acceptance
and usage.
In addition to the future research investigating GRA, it is suggested that the
conceptual decision-making model presented in Chapter II be further evaluated in various
decision-making scenarios, but particularly geospatial decision-making. For example,
future extensions should address geospatial relationships in addition to proximity
183
relationships. Furthermore, in addition to simple point data, as was used in this study, line
and polygon data could provide a richer and more realistic problem solving experiment.
Additionally, this dissertation only addresses individual perceptions related to SDSS
decision-making performance. While this is suitable for most consumer geospatial
decision-making, many business and governmental geospatial decision-making require
group participation.
184
REFERENCES
Abad, M., D az, I., & Vigo, M. (2010). Acceptance of mobile technology in hedonic
scenarios. In Proceedings of the 24th BCS Interaction Specialist Group
Conference, (pp. 250-258). British Computer Society.
Africa, C. (2013). Making cities more liveable with GIS. FutureGov. Retrieved from
http://www.futuregov.asia/articles/2013/jul/30/making-cities-more-liveable-gis/.
Agrawala, M., & Stolte, C. (2001). Rendering effective route maps: improving usability
through generalization. In Proceedings of the 28th
Annual Conference on
Computer Graphics and Interactive Techniques (pp. 241-249). ACM.
Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In J. Kuhl &
J. Beckmann (Eds.), Action-control: From cognition to behavior (pp. 11-39). New
York: Springer-Verlag.
Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior.
Englewood Cliffs, NJ: Prentice-Hall.
Albert, W. S., & Golledge, R. G. (1999). The use of spatial cognitive abilities in
geographical information systems: The map overlay operation. Transactions in
GIS, 3(1), 7-21.
Al-Momani, K., & Noor, N. A. M. (2009). E-Service quality, ease of use, usability and
enjoyment as antecedents of E-CRM performance: An empirical investigation in
Jordan mobile phone services. The Asian Journal of Technology Management.
2(2), 50-63.
American Express (2013). Express Cash ATM Finder. Retrieved from
http://amex.via.infonow.net/locator/cash/
Amoako-Gyampah, K. (2007). Perceived usefulness, user involvement and behavioral
intention: An empirical study of ERP implementation. Computers in Human
Behavior, 23(3), 1232-1248.
Amoako-Gyampah, K., & Salam, A. F. (2004). An extension of the technology
acceptance model in an ERP implementation environment. Information &
Management, 41(6), 731-745.
Anderson, C. L., & Agarwal, R. (2010). Practicing safe computing: A multimethod
empirical examination of home computer user security behavioral intentions. MIS
Quarterly, 34(3), 613-643.
Apple. (2013). Find my friends! Retrieved from
185
http://www.apple.com/icloud/features/find-my-friends.html
Arning, K., & Ziefle, M. (2007). Barriers of information access in small screen device
applications: The relevance of user characteristics for a transgenerational design,
In Universal Access in Ambient Intelligence Environments (pp. 117-136). Berlin
Heidelberg: Springer.
Audet, R. H., & Abegg, G. L. (1996). Geographic information systems: implications for
problem solving. Journal of Research in Science Teaching, 33(1), 21-45.
Bagozzi, R. P., Yi, Y., & Phillips, L. W. (1991). Assessing construct validity in
organizational research, Administrative Science Quarterly, 36, 421-458.
Balog, A. (2011). Testing a multidimensional and hierarchical quality assessment model
for digital libraries. Studies in Informatics and Control, 20(3), 233-246.
Balog, A., & Pribeanu, C. (2010). The role of perceived enjoyment in the students’
acceptance of an augmented reality teaching platform: A structural equation
modeling approach. Studies in Informatics and Control, 19(3), 319-330.
Beemer, B. A., & Gregg, D. G. (2010). Dynamic interaction in knowledge based systems:
An exploratory investigation and empirical evaluation. Decision Support Systems,
49(4), 386-395.
Bing Maps (2013). Retrieved from http://www.bing.com/maps.
Brody, H., Rip, M. R., Vinten-Johansen, P., Paneth, N., & Rachman, S. (2000). Map-
making and myth-making in Broad Street: the London cholera epidemic, 1854.
The Lancet, 356(9223), 64-68.
Brown, S. A., & Venkatesh, V. (2005). Model of adoption of technology in households:
A baseline model test and extension incorporating household life cycle. MIS
Quarterly, 29(3), 399-426.
Brusset, X. (2012). Supply chains: agile, robust or both? In Colloquium on European
Retail Research Book of Proceedings, 65-101.
Campbell, D. J. (1988). Task complexity: A review and analysis. Academy of
Management Review, 13(1), 40-52.
Campbell, S. G. (2011). Users' spatial abilities affect interface usability outcomes
(Doctoral dissertation), University of Maryland, College Park.
Chen, L. D., Gillenson, M. L., & Sherrell, D. L. (2002). Enticing online consumers: an
extended technology acceptance perspective. Information & Management, 39(8),
705-719.
186
Chen, L.D., & Tan, J. (2004). Technology adaptation in E-commerce: Key determinants
of virtual stores acceptance. European Management Journal, 22(1), 74-86.
Chen, R. J. C. (2007). Geographic information systems (GIS) applications in retail
tourism and teaching curriculum. Journal of Retailing and Consumer Services,
14(4), 289-295.
Cherney, I. D., Brabec, C. M., & Runco, D. V. (2008). Mapping out spatial ability: Sex
differences in way-finding navigation. Perceptual and Motor Skills, 107(3), 747-
760.
Chesney, T. (2006). An acceptance model for useful and fun information systems.
Human Technology: An Interdisciplinary Journal on Humans in ICT
Environments, 2(2), 225-235.
Childers, T. L., Carr, C. L., Peck, J., & Carsons, S. (2002). Hedonic and utilitarian
motivations for online retail shopping behavior. Journal of Retailing, 77(4), 511-
535.
Chin, W. W. (1998). The partial least squares approach to structural equation modeling.
In G. A. Marcoulides (Ed.). Modern methods for business research (pp. 295-336).
Mahwah, NJ: Laurence Erlbaum Associates.
Chin, W. W., Thatcher, J. B., & Wright, R. T. (2012). Assessing common method bias:
problems with the ULMC technique. MIS Quarterly, 36(3), 1003-1019.
Clementini, E., Di Felice, P., & van Oosterom, P. (1993). A small set of formal
topological relationships suitable for end-user interaction. In D. Abel, B. C. Ooi
(Eds.), Advances in Spatial Databases: Third International Symposium, SSD '93
Singapore, Proceedings, Lecture Notes in Computer Science, 692(1993), 277-295.
Compeau, D., Marcolin, B., Kelley, H., & Higgins, C. (2012). Generalizability of
information systems research using student subjects – a reflection on our practices
and recommendations for future research, Information Systems Research, 23(4),
1093-1109.
ComScore (2011). U.S. mobile map audience grows 39 percent in past year as fixed-
internet map audience softens slightly, Retrieved from
http://www.comscore.com/Press_Events/Press_Releases/2011/7/U.S._Mobile_Ma
p_Audience_Grows_39_Percent_in_Past_Year
Conroy, M. M., & Gordon, S. I. (2004). Utility of interactive computer-based materials
for enhancing public participation. Journal of Environmental Planning and
Management, 47(1), 19-33.
Conway, J. M., & Lance, C. E. (2010). What reviewers should expect from authors
187
regarding common method bias in organizational research, Journal of Business
and Psychology, 25(3), 325-334.
Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests,
Psychometrika. 16(3), 297-334.
Crossland, M. D., & Wynne, B. E. (1994). Measuring and testing the effectiveness of a
spatial decision support system. In Proceedings of the 27th Annual Hawaii
International Conference on System Sciences, (Volume IV, pp. 542-551). Los
Alamitos, CA: IEEE Computer Society Press.
Crossland, M. D., Wynne, B. E., & Perkins, W. C. (1995). Spatial decision support
systems: An overview of technology and a test of efficacy. Decision Support
Systems, 14(3), 219-235.
Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of
information technology. MIS Quarterly, 13(3), 319-340.
Davis, F. D. (1993). User acceptance of information technology: system characteristics,
user perceptions and behavioral impacts. International Journal of Man-Machine
Studies, 38(3), 475-487.
Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer
technology: a comparison of two theoretical models. Management Science, 35(8),
982-1003.
Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1992). Extrinsic and intrinsic motivation
to use computers in the workplace, Journal of Applied Social Psychology, 22(14),
1111-32.
Dennis, A. R., & Carte, T. A. (1998). Using geographical information systems for
decision making: extending cognitive fit theory to map-based presentations.
Information Systems Research, 9(2), 194-203.
Densham, P. J. (1991). Spatial decision support systems. In D. J. Maguire, M. F.
Goodchild, and D. W. Rhind (Eds.) Geographical Information Systems:
Principles and Applications, (pp. 403-412), New York: John Wiley and Sons.
Department of the Interior (2012). Twitter Earthquake Detector. Retrieved from
http://recovery.doi.gov/press/us-geological-survey-twitter-earthquake-detector-
ted/
Devaraj, S., Easley, R. F., & Crant, J. M. (2008). Research note—how does personality
matter? Relating the five-factor model to technology acceptance and use.
Information Systems Research, 19(1), 93-105.
188
Diamantopoulos, A., Reifler, P., & Roth, K. P. (2008). Advancing formative
measurement models, Journal of Business Research, 61(12), 1203 - 1218.
Djamasbi, S., Strong, D. M., & Dishaw, M. (2010). Affect and acceptance: Examining
the effects of positive mood on the technology acceptance model. Decision
Support Systems, 48(2), 383-394.
Eisenberg, T. A., & McGinty, R. L. (1977). On spatial visualization in college students,
Journal of Psychology, 95(1), 99-104.
Elliott, P., Wakefield, J. C., Best, N. G., & Briggs, D. J. (2000). Spatial Epidemiology:
Methods and Applications, Oxford University Press.
Erskine, M. A., & Gregg, D. G. (2011). Geospatial reasoning ability of business decision
makers: construct definition and measurement. In Proceedings of the Seventeenth
Americas Conference on Information Systems, Detroit, Michigan.
Erskine, M. A., & Gregg, D. G. (2012). The effects of geospatial website attributes on
eImage: An exploratory study. In Proceedings of the 2012 International
Conference on Information Resources Management Information, Vienna, Austria.
Erskine, M. A., & Gregg, D. G. (2013). Impact of geospatial reasoning ability and
perceived task-technology fit on decision-performance: the moderating role of
task characteristics. In Proceedings of the Nineteenth Americas Conference on
Information Systems, Chicago, Illinois.
ESRI, (2013). Retrieved from http://www.esri.com.
Evans, G. W. (1980). Environmental cognition. Psychological Bulletin, 88(2), 259-287.
Fagan, M. H., Neill, S., & Wooldridge, B. R. (2008). Exploring the intention to use
computers: An empirical investigation of the role of intrinsic motivation, extrinsic
motivation, and perceived ease of use. Journal of Computer Information Systems,
48(3), 31-37.
Federal Geographic Data Committee (1999). NSDI Community Demonstration Projects.
Retrieved from
http://www.fgdc.gov/nsdi/library/factsheets/documents/demofactsheet.pdf
Ferguson, R. B. (2012). Location analytics: bringing geography back. MIT Sloan
Management Review. Retrieved from http://sloanreview.mit.edu/feature/location-
analytics-bringing-geography-back/
Fishbein, M. and Ajzen, I. (1975). Belief, attitude, intention, and behavior: an
introduction to theory and research. Reading, Massachusetts: Addison-Wesley.
189
Flanagin, A. J., & Metzger, M. J. (2008). The credibility of volunteered geographic
information. GeoJournal, 72(3-4), 137-148.
Fornell, C., & Larcker, D. (1981) Evaluating Structural Equation Models with
unobservable variables and measurement error. Journal of Marketing Research,
18(1), 39-50.
Frownfelter-Lohrke, C. (1998). The effects of differing information presentations of
general purpose financial statements on users’ decisions. Journal of Information
Systems, 12(2), 99-107.
Gefen, D., Karahanna, D., & Straub, D. W. (2003). Trust and TAM in online shopping:
an integrated model. MIS Quarterly, 27(1), 51-90.
Gefen, D., & Straub, D., (2005). A practical guide to factorial validity using PLS-Graph:
tutorial and annotated example, Communications of the Association for
Information Systems, 16(1), 91-109.
Gefen, D., & Straub, D. W. (1997). Gender differences in the perception and use of e-
mail: an extension to the technology acceptance model. MIS Quarterly, 21(4),
389-400.
Gerow, J. E., Ayyagari, R., Thatcher, J. B., & Roth, P. L. (2013). Can we have fun @
work? The role of intrinsic motivation for utilitarian systems. European Journal
of Information Systems, 22(3), 360-380.
Gilbert, D., Lee-Kelley, L., & Barton, M. (2003). Technophobia, gender influences and
consumer decision-making for technology-related products. European Journal of
Innovation Management, 6(4), 253-263.
Gill, T. G., & Hicks, R. C., (2006). Task complexity and informing science: a synthesis.
Informing Science Journal, 9, 1-30.
Gliem, J. A., & Gliem R. R. (2003). Calculating, interpreting, and reporting Cronbach’s
alpha reliability coefficient for Likert-type scales. In Proceedings of the 2003
Midwest Research to Practice Conference in Adult, Continuing and Community
Education, Columbus, OH.
Goodchild, M. (2007) Citizens as sensors: The world of volunteered geography.
GeoJournal, 69, 211-221.
Goodhue, D. L. (1995). Understanding user evaluations of information systems.
Management Science, 41(12), 1827-1844.
Goodhue, D. L., & Thompson, R. L. (1995). Task-technology fit and individual
performance. MIS Quarterly, 19(2), 213-236.
190
Google Maps (2013). Retrieved from http://maps.google.com
Griggs, B. (2013). New Google Maps can help you avoid traffic, CNN International,
Retrieved from http://www.cnn.com/2013/08/20/tech/mobile/google-waze-
mobile-maps.
Grimshaw, D.J. (2001). Harnessing the power of geographical knowledge: the potential
for data integration in an SME. International Journal of Information
Management, 21, 183-191.
Gundotra, V. (2010). To 100 million and beyond with Google Maps for mobile, Google
Mobile Blog, Retrieved from http://googlemobile.blogspot.com/2010/08/to-100-
million-and-beyond-with-google.html
Hair Jr., J. F., Black, W. C., Babin, B, J., & Anderson, R. E. (2010). Multivariate data
analysis. Seventh edition. USA: Prentice Hall.
Hair Jr., J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt, M. (2014). A primer on partial
least squares structural equation modeling (PLS-SEM), Los Angeles, California:
SAGE Publications.
Hair Jr., J. F., Ringle, C. M., & Sarstedt, M. (2011). PLS-SEM: indeed a silver bullet, J.
of Marketing Theory and Practice, 19, 139-151.
Herath, T., & Rao, H. R. (2009). Protection motivation and deterrence: A framework for
security policy compliance in organizations, European Journal of Information
Systems, 18(2), 106-125.
Herskovits, A. (1998). Schematization. In P. Olivier & K. P. Gapp (Eds.), Representation
and processing of spatial expressions pp. 149-162. Mahwah, New Jersey:
Lawrence Erlbaum Associates.
Hess, R. L., Rubin, R. S., & West, L. A. (2004). Geographic information systems as a
marketing information system technology. Decision Support Systems, 38(2), 197-
212.
Ho, S. Y. (2010). The effects of location-based mobile personalization on users’
behavior, In Proceedings of Pacific Asia Conference on Information Systems,
Taipei, Taiwan.
Hollenbeck, J. R., & Klein, H. J. (1987). Goal commitment and the goal-setting process:
problems, prospects, and proposals for future research. Journal of Applied
Psychology, 72(2), 212-220.
Hu, P. J., Chau, P. Y. K., Sheng, O. R. L., & Tam, K. Y. (1999). Examining the
191
technology acceptance model using physician acceptance of telemedicine
technology. Journal of Management Information Systems, 16(2), 91-112.
Huang, M. -H. (2000). Information load: its relationship to online exploratory and
shopping behavior. International Journal of Information Management, 20(5),
337-437.
Hung, S. -Y., Ku, Y. -C., Ting-Peng, L., & Chang-Jen, L. (2007). Regret avoidance as a
measure of DSS success: An exploratory study. Decision Support Systems, 42(4),
2093-2106.
Hwang, J., Jung, J., & Kim, G. J., (2006). Hand-held virtual reality: a feasibility study, In
Proceedings of ACM Virtual Reality Software and Technology (VRST06),
Limassol, Cyprus, 356-363.
Ibrahim, A.M. (2001). Differential responding to positive and negative items: the case of
a negative item in a questionnaire for course and faculty evaluation.
Psychological Reports, 88(2), 497-500.
Ives, B. (1982). Graphical user interfaces for business information systems. MIS
Quarterly, 6, 1982, 15-47.
Jankowski, P., & Nyerges, T. (2001). GIS-supported collaborative decision-making:
Results of an experiment. Annals of the Association of American Geographers,
91(1), 48-70.
Jarupathirun, S., & Zahedi, F. (2001). A theoretical framework for GIS-based spatial
decision support systems: utilization and performance evaluation, In Proceedings
of the Americas Conference on Information Systems 2001, 245-248.
Jarupathirun, S., & Zahedi, F. M. (2007). Exploring the influence of perceptual factors in
the success of web-based spatial DSS. Decision Support Systems, 43(3), 933-951.
Jarvenpaa, S. L. (1989). The effect of task demands and graphical format on information
processing strategies, Management Science, 25(3), 285-303.
Karahanna, E., Straub, D. W., & Chervany, N. L. (1999). Information technology
adoption across time: a cross-sectional comparison of pre-adoption and post-
adoption beliefs. MIS Quarterly, 23(2), 183-213.
Karimi, J., Somers, T. M., & Gupta, Y. P. (2004). Impact of environmental uncertainty
and task characteristics on user satisfaction with data. Information Systems
Research, 15(2), 175-193.
Kelly, G. C., Tanner, M., Vallely, A., & Clements, A. (2012). Malaria elimination:
moving forward with spatial decision support systems. Trends in Parasitology,
192
28(7), 297-304.
Kelton, A. S., Pennington, R. R., & Tuttle, B. M. (2010). The effects of information
presentation format on judgment and decision making: a review of the
information systems research. Journal of Information Systems, 24(2), 79-105.
Kerski, J. J. (2000). The implementation and effectiveness of geographic information
systems technology and methods in secondary education. (Unpublished doctoral
dissertation), University of Colorado, Boulder.
Khalili, N., Wood, J., & Dykes, J. (2009). Mapping geography of social networks, In D.
Fairbairn (Ed.), Proceedings of the GIS Research UK 17th Annual Conference,
University of Durham, Durham, UK, 311-315.
Khalili, N., Wood, J., & Dykes, J. (2010). Analyzing uncertainty in home location
information in a large volunteered geographic information database, In
Proceedings of the GIS Research UK 18th Annual Conference, University
College London, London, UK, 57-63.
Klein, J., Moon, Y., & Picard, R. W. (2002). This computer responds to user frustration:
theory, design, and results. Interacting with Computers, 14(2), 119-140.
Klippel, A., Richter, K. -F., Barkowsky, T., & Freksa, C. (2005). The cognitive reality of
schematic maps. In L. Meng, A. Zipf, & T. Reichenbacher (Eds.) Map-based
Mobile Services - Theories, Methods and Implementations. Springer, Berlin, 57-
74.
Kozlowski, L. T, & Bryant, K. J. (1977). Sense of direction, spatial orientation, and
cognitive maps. Journal of Experimental Psychology: Human Perception and
Performance, 3(4), 590-598.
Larson, K., & Czerwinski, M. (1998). Web page design: implications of memory,
structure and scent for information retrieval. In Proceedings of the SIGCHI
Conference on Human Factors in Computing Systems, ACM Press/Addison-
Wesley Publishing Co, 25-32.
Lavrinc, D. (2013). Google needs to make maps for motorcycles. Wired. Retrieved from
http://www.wired.com/autopia/2013/11/google-maps-for-motorcyclists/
Lawton, C. A. (1994). Gender differences in way-finding strategies: relationship to
spatial ability and spatial anxiety. Sex Roles, 30(11/12), 765-779.
Lee, A. S., & Baskerville, R. L. (2003). Generalizing generalizability in information
systems research. Information Systems Research, 14(3), 221-243.
Lee, A. S., & Baskerville, R. L. (2012). Conceptualizing generalizability: new
193
contributions and a reply, MIS Quarterly, 36(3), 749-761.
Lee, J., & Bednarz, R. (2009). Effect of GIS learning on spatial thinking. Journal of
Geography in Higher Education, 33(2), 183-198.
Lee, H. -H., Fiore A. M., & Kim, J. (2006). The role of the technology acceptance model
in explaining effects of image interactivity technology on consumer responses.
International Journal of Retail & Distribution Management, 34(8), 621-644.
Lee, A. S., & Hubona, G. S. (2009). A scientific basis for rigor in information systems
research. MIS Quarterly, 33(2), 237-262.
Legris, P., Ingham, J.,& Collerette, P. (2003). Why do people use information
technology? A critical review of the technology acceptance model. Information &
Management, 40(3), 191-204.
Lei, P -L. Kao, G. Y -M, Lin, S. S. J., & Sun, C -T. (2009). Impacts of geographical
knowledge, spatial ability and environmental cognition on image searches
supported by GIS software. Computers in Human Behavior, 25(6), 1270-1279.
Liang, H., Saraf, N., Hu, Q., & Xue, Y. (2007). Assimilation of enterprise systems: the
effect of institutional pressures and the mediating role of top management. MIS
Quarterly, 31(1), 59-87.
Liker, J. K., & Sindi, A. A. (1997). User acceptance of expert systems: a test of the
theory of reasoned action, Journal of Engineering and Technology Management,
14(2), 147-173.
Likert, R. (1932). A technique for the measurement of attitudes. Archives of Psychology,
22, 140, 1-55.
Lin, J. C., & Lu, H. (2000). Towards an understanding of the behavioral intention to use a
web site, International Journal of Information Management, 20(3), 197-208.
Lindell, M. K., & Whitney, D. J. (2001). Accounting for common method variance in
cross-sectional research designs. Journal of Applied Psychology, 86(1), 114-121.
Linn M. C., & Petersen, A. C. (1985). Emergence and characterization of sex differences
in spatial ability: a meta-analysis. Child Development, 56(6), 1479-1498.
Liu, L., Li, C., & Zhu, D. (2012). A new approach to testing nomological validity and its
application to a second-order measurement model of trust. Journal of the
Association for Information Systems, 13(12), 950-975.
Luo, X., Li, H., Zhang, J., & Shim, J. P. (2010). Examining multi-dimensional trust and
multi-faceted risk in initial acceptance of emerging technologies: an empirical
194
study of mobile banking services. Decision Support Systems, 49(2), 222-234.
MacKenzie, S. B., Podsakoff, P. M., & Jarvis, C. B. (2005). The problem of measurement
model misspecification in behavioral and organizational research and some
recommended solutions. Journal of Applied Psychology, 90(4), 710-730.
MacKenzie, S. B., Podsakoff, P. M., & Podsakoff, N. P. (2011). Construct measurement
and validation procedures in MIS and behavioral research: integrating new and
existing techniques. MIS Quarterly, 35(2), 293-334.
Malczewski, J. (2006). GIS-based multicriteria decision analysis: a survey of the
literature. International Journal of Geographical Information Science, 20(7), 703-
726.
Mäntylä, T. (2013). Gender differences in multitasking reflect spatial ability.
Psychological Science, 24(4), 514-520.
Marakas, G. M., Johnson, R. D., & Clay, P. F. (2007). The evolving nature of the
computer self-efficacy construct: an empirical investigation of measurement
construction, validity, reliability and stability over time. Journal of the
Association for Information Systems, 8(1), 16-46.
MasterCard (2013). ATM Locator, Retrieved from http://www.mastercard.us/cardholder-
services/atm-locator.html.
Mathieson, K. (1991). Predicting user intentions: comparing the technology acceptance
model with the theory of planned behavior. Information Systems Research, 2(3),
173-191.
Mehrabian, A., & Russell, J. A. (1974). An Approach to Environmental Psychology.
Cambridge, MA: The MIT Press, 88-94.
Meilinger, T., & Knauff, M. 2008. Ask for directions or use a map: a field experiment on
spatial orientation and wayfinding in an urban environment, Journal of Spatial
Science, 53(2), 13-23.
Mennecke, B. E. (1997). Understanding the role of geographic information technologies
in business: applications and research directions. Journal of Geographic
Information and Decision Analysis, 1(1), 44-68.
Mennecke, B.E., & Crossland, M.D. (1996). Geographic information systems:
applications and research opportunities for information systems researchers. In
Proceedings of the 29th Hawaii International Conference on System Sciences,
537-546.
Mennecke, B. E., Crossland, M. D., & Killingsworth, B. L. (2000). Is a map more than a
195
picture? The role of SDSS technology, subject characteristics, and problem
complexity on map reading and problem solving. MIS Quarterly, 24(4), 601-629.
Meyer, J. W., Butterick, J., Olkin, M., & Zack, G. (1999). GIS in the K–12 curriculum: a
cautionary note, Professional Geographer, 51(4), 571–578.
Miyake, A., Friedman, N. P., Rettinger, D. A., Shah, P., & Hegarty, M. (2001). How are
visuospatial working memory, executive functioning, and spatial abilities related?
A latent-variable analysis. Journal of Experimental Psychology: General, 130(4),
621-640.
Mook, J., Kleijn, W. C, & van der Ploeg, H.M. (1992). Positively and negatively worded
items in a self-report measure of dispositional optimism, Psychological Reports,
71(1), 275-278.
Noble, D. J. (2012). Predicting the epidemic: a study of diabetes risk profiling in a multi-
ethnic inner city population (Doctoral dissertation).
Novak, J. & Schmidt, S. (2009). When joy matters: the importance of hedonic stimulation
in collocated collaboration with large-displays, In Proceedings of Interact 2009,
Uppsala, Sweden.
O’Cass, A. & Fenech, T. (2003). Web retailing adoption: Exploring the nature of internet
users web retailing behavior. Journal of Retailing and Consumer Services, 10(2),
81-94.
Olsen, T. P. (2000). Situated student learning and spatial informational analysis for
environmental problem. (Unpublished doctoral dissertation). University of
Wisconsin-Madison, Madison.
Ozimec, A. M., Natter, M., & Reutterer, T. (2010). Geographical information systems-
based marketing decisions: effects of alternative visualizations on decision
quality. Journal of Marketing, 74(6), 94-110.
Payne, J. W. (1976). Task complexity and contingent processing in decision-making: an
information search and protocol analysis, Organizational Behavior & Human
Performance, 16(2), 366-387.
Pick, J. B. (2004). Geographic information systems: a tutorial and introduction.
Communications of the Association for Information Systems, 14(1), 307-331.
Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common
method biases in behavioral research: a critical review of the literature and
recommended remedies. Journal of Applied Psychology, 88(5), 879-903.
Polites, G. L., Roberts, N., & Thatcher, J. (2012). Conceptualizing models using
196
multidimensional constructs: a review and guidelines for their use. European
Journal of Information Systems, 21(1), 22-48.
Rafi, A., Anuar, K., Samad, A., Hayati, M., & Mahadzir, M. (2005). Improving spatial
ability using a Web-based Virtual Environment (WbVE). Automation in
Construction, 14(6), 707-715.
Rasaf, M. R., Ramezani, R., Mehrazma, M., Rasaf, M. R. R., & Asadi-Lari, M. (2012).
Inequalities in cancer distribution in Tehran; A Disaggregated Estimation of 2007
Incidencea by 22 Districts, International Journal of Preventive Medicine, 3(7),
483.
Reiterer, H., Mann, T. M., Mußler, G., & Bleimann, U. (2000). Visualisierung von
entscheidungsrelevanten daten für das management, HMD, Praxis der
Wirtschaftsinformatik, 212, 71 - 83.
Ringle, C. M., Götz, O, Wetzels, M., & Wilson, B. (2009). On the use of formative
measurement specifications in structural equation modeling: a Monte Carlo
Simulation study to compare covariance-based and partial least squares model
estimation methodologies. In Research Memoranda from Maastricht (METEOR)
Ringle, C. M., Wende, S., & Will, S. (2005). SmartPLS (M3) Beta, Hamburg, Retrieved
from http://www.smartpls.de.
Re/Max (2013). Property search. Retrieved from http://www.remax.com/advancedsearch/
Rogers, E. M. (1983). Diffusion of Innovations (3rd
ed.), New York: The Free Press.
Rusch, M. L., Nusser, S. M., Miller, L. L., Batinov, G. I., & Whitney, K. C. (2012).
Spatial ability and map-based software applications. In Proceedings of the Fifth
International Conference on Advances in Computer-Human Interactions, 35-40.
Schmitt, N., & Stults, D. M. (1985). Factors defined by negatively keyed items: the result
of careless respondents?, Applied Psychological Measurement, 9(4), 367-373.
Seaman V. (1798). An inquiry into the cause of the prevalence of yellow fever in New
York, (1st ed.), New York: The Medical Repository.
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-
experimental designs for generalized causal inference, Boston: Houghton Mifflin
Company.
Sieber, R. (2006). Public participation geographic information systems: a literature
review and framework, Annals of the Association of American Geographers,
96(3), 491-507.
197
Sirola, M. (2003). Decision concepts. In IEEE International Workshop on Intelligent
Data Acquisition and Advanced Computing Systems: Technology and
Applications, 59-62.
Skupin, A., & Fabrikant, S. I. (2003). Spatialization methods: a cartographic research
agenda for non-geographic information visualization, Cartography and
Geographic Information Science, 30(2), 99-119.
Slocum, T. A., Blok, C., Jiang, B., Koussoulakou, A., Montello, D. R., Fuhrmann, S., &
Hedley, N. R. (2001). Cognitive and usability issues in geovisualization.
Cartography and Geographic Information Science, 28(1), 61-75.
Smelcer, J. B., & Carmel, E. (1997). The effectiveness of different representations for
managerial problem solving: comparing tables and maps. Decision Sciences,
28(2), 391-420.
Snow, J. (1849). On the mode of communication of cholera (1st ed.), London: J.
Churchill.
Snow, J. (1855). On the mode of communication of cholera (2nd
ed.), London: J.
Churchill.
Spector, P. E., Van Katwyk, P. T., Brannick, M. T., & Chen, P. Y. (1997). When two
factors don’t reflect two constructs: how item characteristics can produce
artifactual factors, Journal of Management, 23(5), 659-677.
Speier, C. (2006). The influence of information presentation formats on complex task
decision-making performance. International Journal of Human-Computer
Studies, 64(11), 1115-1131.
Speier, C., & Morris, M. G. (2003). The influence of query interface design on decision-
making performance. MIS Quarterly, 27(3), 397-423.
Strohecker, C. (2000). Cognitive zoom: from object to path and back again, In Spatial
Cognition II, Berlin: Springer Verlag.
Subramanian, G. H. (1994). A replication of perceived usefulness and perceived ease of
use measurement. Decision Sciences, 25(5/6), 863-874.
Subsorn, P., & Singh, K. (2007). DSS applications as a business enhancement strategy, In
Proceedings from the 3rd annual Transforming Information and Learning
Conference.
Swink, M., & Speier, C. (1999). Presenting geographic information: effects on data
aggregation, dispersion, and users’ spatial orientation. Decision Sciences, 30(1),
169-196.
198
Taylor, S., & Todd, P. A. (1995). Understanding information technology usage: a test of
competing models. Information Systems Research, 6(2), 144-176.
Thong, J. Y., Hong, S. J., & Tam, K. Y. (2006). The effects of post-adoption beliefs on
the expectation-confirmation model for information technology continuance.
International Journal of Human-Computer Studies, 64(9), 799-810.
T-Mobile (2013). Check your coverage. Retrieved from http://www.t-
mobile.com/coverage.html
Tonkin, T. (1994) Business geographics impacts corporate America. Business
Geographics, 2(2), 27-28.
Tourangeau, R., Rips, L. J., & Rasinski, K. (2000). The psychology of survey response.
Cambridge, England: Cambridge University Press.
Tsang, E. W., & Williams, J. N. (2012). Generalization and induction: misconceptions,
clarifications, and a classification of induction. MIS Quarterly, 36(3), 729-748.
Tversky, B., & Lee, P. U. (1998). How space structures language, In C. Freksa, C. Habel
& K. F. Wender (eds.) Spatial Cognition: An interdisciplinary approach to
representation and processing of spatial knowledge (pp. 157-175), Berlin:
Springer Verlag.
van der Heijden, H. (2004). User acceptance of hedonic information systems, MIS
Quarterly, 28(4), 695-704.
Velez, M. C., Silver, D., & Tremaine, M. (2005). Understanding visualization through
spatial ability differences, In Proceedings of IEEE Visualization (pp. 23-28),
Minneapolis, MN, USA.
Venkatesh, V., & Davis, F. D. (2000). A theoretical extension of the technology
acceptance model: four longitudinal case studies. Management Science, 46(2),
186-204.
Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of
information technology: toward a unified view, MIS Quarterly, 27(2), 425-278.
Verkasalo, H., López-Nicolás, C., Molina-Castillo, F. J., & Bouwman, H. (2010).
Analysis of users and non-users of smartphone applications. Telematics and
Informatics, 27(3), 242-255.
Versace, C. (2013) Mapping heats up as Apple buys Embark, Google integrates Waze…
What’s next? Forbes, Retrieved from
http://www.forbes.com/sites/chrisversace/2013/08/26/mapping-heats-up-as-apple-
199
buys-embark-google-integrates-waze-whats-next/
Vessey, I. (1991). Cognitive fit: a theory-based analysis of the graphs vs. tables literature.
Decision Sciences, 22(2), 219-240.
Vijayasarathy, L.R. (2004). Predicting consumer intentions to use on-line shopping: the
case for an augmented technology acceptance model. Information &
Management, 41(6), 747-762.
Vinzi, V. E., Chin, W. W., Henseler, J., & Wang, H. (2010). Handbook of partial least
squares: Concepts, methods and applications. Berlin: Springer.
Vlachos, P. A., & Theotokis, A. (2009). Formative versus reflective measurement for
multidimensional constructs. Social Science Research Network, Retrieved from
http://ssrn.com/abstract=1521095
Whitney, K. C., Batinov, G. J., Miller, L. L., Nusser, S. M., & Ashenfelter, K. T. (2011).
Exploring a map survey task’s sensitivity to cognitive ability, In The Fourth
International Conference on Advances in Computer-Human Interactions (pp. 63-
68), Gosier, Guadeloupe, France.
Winn, W., Hoffman, H., Hollander, A., Osberg, K., Rose, H., & Char, P. (1997). The
effect of student construction of virtual environments on the performance of high-
and low-ability students, In Annual Meeting of the American Educational
Research Association, Chicago.
Yang, C., Raskin, R., Goodchild, M., & Gahegan, M. (2010). Geospatial
cyberinfrastructure: past, present and future. Computers, Environment and Urban
Systems, 34(4), 264-277.
Yang, C., Wong, D. W., Yang, R., & Li, Q. (2005). Performance-improving techniques in
web-based GIS, International Journal of Geographical Information Science,
19(3), 319-342.
Yi, M.Y., Jackson, J. D., Park, J. S., & Probst, J. C. (2006). Understanding information
technology acceptance by individual professionals? Toward an integrative view.
Information & Management, 43(3), 350-363.
Zigurs, I., & Buckland, B. (1998). A theory of task/technology fit and group support
systems effectiveness, MIS Quarterly, 22(3), 313-344.
Zillow (2013). Retrieved from http://www.zillow.com/mobile/.
Zipf, A. (2002). User-adaptive maps for location-based services (LBS) for tourism. In
Proceedings of the 9th International Conference for Information and
Communication Technologies in Tourism, ENTER 2002, Innsbruck, Austria.
