An Open Spatial Systems Framework for Place-Based

Decision-MakingMarynia Kolak, PhD Candidate

PH-GIS | April 14, 2016

Background● A push towards place-based policy

● Need for distilling causal links

● Complex spatial organization & processes

● New types Big Data

● Limitations of existing infrastructures

Chicago Boundaries (Bill Rankin 2009) 2

StreamingBigDynamic

StaticSmallRigid

Absence of an integrated systems framework means that spatial effects are not considered or accounted for effectively or consistentlyin a decision-making environment.

3

Issues from Ignoring Spatial EffectsTheoretical: may confuse process being studied, miss

important signals, & violate core assumptions.

Methodological: may skew and/or bias results that

impact decision-making.

Technological: analysis and/or dynamic decision-

making may not be feasible, especially in a Big Data

context.

In a policy or decision-making context, integrating all three components is essential for on-the-fly analysis of spatially dynamic phenomena.

5

Space serves as the Place for IntegrationA spatial framework for distilling

causal links for policy &

intervention assessment

considers a more comprehensive

understanding of how spatial

effects impact the data

generating process.

theoretical

methodological technologicalinfrastructure design

research design usability

designSPATIAL

DATASCIENCE

forDECISION-MAKING

6

Towards an (Open) Spatial Technological FrameworkHow can we better integrate data and spatial analysis in a flexible framework?

● Challenges with data sourcing, processing, integration, and ownership

● Need for sharing results in an easy, updateable cross-platform way

● Gap between web development and spatial analytic communities

● Adding not just visualization, but analytics to an open web environment

Desktop Environments: ArcGIS, QGIS, GeoDa, PostGIS

Web Environments: ArcGIS Explorer, Google Products, Policy Map, (GeoDa),

Open, Customizable Web Environments: HTML/CSS, Javascript, (Python), & much more...

Open Source EnvironmentsPros:

Software is free

Code is shared, and can be improved over time (ie.

when bugs emerge, community fixes them)

Project can be customized and improved

Cons:

Steep learning curve: learning or updating

programming skills to make and fix code

If a project doesn’t have a large community, bugs

won’t be fixed (unless you try to)

Open Source projects thrive where there is:● Community, collaboration, and team environments● Space, time, and availability to work on improvement● Phased development and innovation are encouraged● Flexible, working connection of multiple components

Project Overview+

Spatial (Data Science) Decision Support Application● Built-in Spatial Framework

● Allows for spatially dynamic processes

● Dynamic development & evaluation of place-based policies

Customized Application Goals:

● Asset Mapping - Supports Needs Assessment

● Identify Areas for Treatment - Based on Risk Assessment

● Evaluate Policies & Interventions - Causal Analytic Framework

10

increasing complexity

Healthy Access, Healthy Regions ProjectCollaborators: Chicago Department of Public Health (Epidemiology, Policy Analysis, and

Innovation Groups); GeoDa Center, Vader Lab, & ChainBuilder Project at Arizona State

University; Goldstein Research at the Harris School of Public Policy at U of C, and more...

Builds on previous work:

● Spatial Data Warehouse Infrastructure (John Hopkins University 2013)

● West Humboldt Park Resource Mapping (WHPD, OLA, Northwestern, 2013)

● Work with CDPH & multiple community organizations since 2012

● Chicago Food Access Inequity from 2007-2014 (Northwestern-led study, 2016)

● Engineers Without Borders principles & lessons learned (volunteer since 2010)

11

Application GoalsCDPH Goal: Evaluate issues of health accessibility across Chicago

Dissertation Goal: Consider spatially dynamic causal links of health access inequity

Phased Development:

● Community Resource Map and Data Exploration (public-facing)

● Health Risk Indicator Map and Data Exploration (public-facing)

● Service Area Network Analysis and Simulation (analyst-facing)

● Data-Driven Regionalization Analysis (analyst-facing)

● Health Accessibility Model Comparison (analyst-facing)

Application Objectives● Visualizes health providers and human services locations by category;

● Calculates service areas for locations along transportation routes;

● Provides descriptive summaries of communities in service areas of key health providers;

● Visualizes distributions of vulnerable and at-risk populations according to demographic and socioeconomic

characteristics;

● Identifies and summarizes areas of need where services are lacking based on demographic and

socioeconomic need, indexed by service category;

● Calculates new targeted zones of risk or vulnerability according to community need and service coverage

attributes; and

● Identifies relationships between the built environment and disease, by category

Data+

● Different types of information stored

● Different ways of storing information

● Sharing data across organizations

● Managing data updates

● Visualizing data on maps

● Shoestring budget for technology

● Limited staff resources

● Domain knowledge with organization

Resource Mapping ChallengesWhat is the ultimate goal, as defined by the community?

What technology is available to (minimally) achieve that goal?

… That meets the needs of the community?

… Five years from now?

● Visualize and Explore Data

● Dynamic and shareable

● Data could be owned, managed, and

updated by community-based

organizations and other partners

Resource Mapping Needs

Develop a shared community resource for the community, by the community.

VGI Resource MappingCase Example: West Chicago

● Community resources consolidated and

shared across two organizations (West

Humboldt Park Development Council and

Diabetes Link) for the West Side of Chicago

● Initiated while at Northwestern University

● Web map developed by customizing open

source code and published online

● Underrepresented talent recruited to support

development (data, front-end coding)

● makosak.github.io/HumboldtResources

Team Effort!

- West Humboldt Park Development Council

- Our Lady of Angels- Northwestern University- NM Hospital- GeoDa Center- Volunteers from GIS

course, Hack Day 2015

VGI Resource MappingChicago-wide VGI Framework

● New collaboration with the Chicago

Department of Public Health

● Health providers used by senior planning

analyst uploaded to dynamic Fusion table

● Web map application developed

● makosak.github.io/ChiHealthAccess

Next Steps:

● Determine best method of uploading data in

easy-to-use, effective environment

● Share with CBO network by CDPH

Existing Web Page:

Connects Health Providers

Future Web Page:

Connects Community Resources from across Chicago?

Show me nearby resources

22http://makosak.github.io/HumboldtResources

flexible framework

dynamic spatial data

Analysis+

Types of Analysis● Choropleth maps with multiple Data Classifications

○ Equal Interval, Standard Deviation, Fisher Jenks, CDPH Categories

○ Javascript breaks and/or PySal assignment

● Service Area Network Algorithm

○ Distance- or time-based

○ Open Street Maps roads (PBF file) and Fusion Table points (as JSON)

○ Feng Wang at the Vader Lab

● Hot Spot/ Cold Spot and Outlier detection

○ LISA Statistics and cluster maps

● Max-P Regions and Cluster Analysis○ Goldstein Research Group open project: “Blobs”

Show me 1-mile service areas for hospitals

25http://makosak.github.io/chihealthaccess/

dynamic exploration

analytic API integration

Simulation ExperimentIn Development:

● Run analysis with a model to

develop a quantifiable outcome.

● Add, remove, or change a point.

● Re-run analysis; compare results.

● Repeat as much as you’d like.

Technology:

● Automated Workflows

● Data and Analysis both act as

services (ie. called by URL)

● Chainbuilder, Open ASU Project

Which unit should I use for analysis?

30http://makosak.github.io/chihealthaccess/

check assumptions

spatial framework forresearch design

Max-P Algorithm (via “Blobs”)Max-P is a type of regionalization algorithm:

● Regions are a group of “areas,” and there are

generally multiple regions in a particular dataset.

● Regions are binned according to the similarities in

(multivariate) data

● Can define number of areas that must make up a

region, as well as min number of persons

● In epidemiology: may need similar baseline regions to

not violate core assumptions of inference

Systems Integration

+

Methods and Tools of IntegrationMethods of Analysis:

ESDA (choropleth maps, relative risk maps, Moran’s I, LISA cluster maps, histograms)

Spatial Econometrics, Causal Methodology Toolbox

Tools of Integration (Data, Analytics, Visualization):

PostGresSQL, POSTGIS, python, javascript, html/CSS, Flask, D3, Leaflet, Chainbuilder,

Google Fusion Tables, Google Map API, Plenar.io API

Usability Testing: surveys, task-based interviews, focus groups, & stakeholder meetings

34

Application FrameworkDevelop a dynamic and flexible

framework to integrate multiple data

types and formats, allow for scaling, allow

for updates.

Join on space and time in a spatially

enabled data system, rather than object

identifiers, when available.

Privilege services (APIs) and dynamic

updates rather than static datasets.

Allow for service-driven visualization and

analytics in final environment.

35

Coming Soon...Timeframe for final completion:

Summer 2017

New datasets, analytics, and resources!

Connecting larger datasets:

- Sensor Networks?

- Social Media?

In July 2016, GeoDa Center moves to the new

Center for Spatial Data Science

AcknowledgementsChicago Department of Public Health

Vader Lab, GeoDa Center, and ChainBuilder at

Agency for Healthcare Research & Quality

GeoDa Fellowship, Arizona State University

Stan Lesny Scholarship, Kosciuszko Foundation

Questions? email mkolak@asu.edu