25
SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

  • View
    216

  • Download
    0

Embed Size (px)

Citation preview

Page 1: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

SOWK 6003 Social Work Research

Week 10 Quantitative Data Analysis

By Dr. Paul Wong

Page 2: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Overview

Introduction

Levels of Measurement

Coding, Data Entry, and Cleaning

Univariate Analysis

Bivariate Analysis

Multivariate Analysis

Descriptive Statistics and Qualitative Research

Page 3: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Introduction

Quantitative AnalysisTechniques used to convert data to a numerical form

Quantifying data is necessary when statistical analyses are desired

Page 4: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Levels of Measurement

An attribute (e.g., male, female) is a characteristic or quality of a variable, and a variable (e.g., sex) are logical sets of attributes

Different types of attributes (e.g., sex, age, social class) represent different levels of measurement: nominal, ordinal, interval, and ratio

Page 5: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Levels of Measurement

A given variable can sometimes be measured at different levels of measurement (e.g., age)

Page 6: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Nominal Measures

Include those variables with only discrete, nonmetric or categorical attributes

In other words, they include variables whose attributes are different from one another (e.g., sex, ethnicity)

Code numbers are assigned to the different attributes or categories of a variable (e.g., “0” = male, “1” = female), but the code numbers have no quantitative meaning

Page 7: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Ordinal Measures

Include those variables whose attributes may be rank-ordered

– E.g., prejudice as composed of very prejudiced, somewhat prejudiced, and not at all prejudiced

Code numbers are assigned to the categories, but the precise differences or distance between the categories is unknown - we only know the order of the categories (e.g., high to low, more to less)

Page 8: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Interval and Ratio Measures

Interval measures include those variables whose attributes are not only rank-ordered but also separated by a uniform distance between them (e.g., IQ)

Ratio measures are the same as interval measures except ratio measures are based on a true zero point (e.g., age)

Page 9: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Coding, Code Categories, and Codebooks

Coding– The goal is the conversion of data items into

numerical codes, necessary for statistical analyses

– Often occurs after the data have been collected using computer programs, such as SPSS

– Coding approaches vary and should be appropriate to the theoretical concepts under investigation

Page 10: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Coding, Code Categories, and Codebooks

Two basic approaches to coding:

– Well-developed coding scheme or categories based on research purpose (e.g., items with pre-determined categories), or

– Codes generated from data as discussed in Chapter 19 (e.g., open-ended items without response categories)

Page 11: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Coding, Code Categories, and Codebooks

Codebooks describe the locations of variables and list their attributes and assigned codes

Codebooks have two primary functions

– Primary guide during the coding process

– Guide for locating variables and interpreting codes during data analysis

Page 12: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Data Entry

Data entry may be approached in a variety of ways, depending on the original form of your data

Data can be entered

– Directly into computer program, such as SPSS or Excel

– Using optical scan sheets

– Using computer-assisted telephone interviewing (CATI) or online surveys

Page 13: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Data Cleaning

After entering the data, the next step is to eliminate error – that is, “clean” the data

Possible-code cleaning involves the process of checking to see that only those codes assigned particular attributes appear in the data files– Some computer programs can check for data

errors

Page 14: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Univariate Analysis

Analysis of a single variable

The original data collected with regard to a single variable are usually difficult, if not impossible, to interpret

Data reduction involves summarizing the original data to make them more manageable

Page 15: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Univariate Analysis

Several techniques are available to make original data more manageable:

– Frequency distributions

– Measures of central tendency:

• Mean: arithmetic mean, or “center or gravity”

• Median: middle attribute in the ranked distribution of attributes

• Mode: most frequent attribute

Page 16: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Univariate Analysis

Several techniques are available to make original data more manageable:

– Measures of dispersion provide a summary of the distribution of cases around some central value

• Range, the distance between the highest and lowest value, is the simplest dispersion measure

• Standard deviation is the most common and is used to get an idea of far away from the mean the values in our data are falling

Page 17: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Univariate Analysis

Measures of central tendency and dispersion should be used for interval or ratio level variables and may not be appropriate for all variables – e.g., discrete variables

However, technical violations are common and may be useful, and caution should be used to avoid misrepresenting something that is not truly precise

Page 18: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

An Example of a Univariate Table

Page 19: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Bivariate Analysis

Examines relationships between two variables, typically for explanatory purposes– Divides cases into subgroups according to their

attributes on some independent variable

– Describes each subgroup in terms of some dependent variable

– Compares the dependent variable descriptions of the subgroups

– Interprets observed differences as statistical associations between the independent and dependent variables

Page 20: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Constructing and Reading Bivariate Tables

Often referred to as contingency tables

Provide clear, succinct table heading

Present original content of variables, if possible, or in the text with a paraphrase in the table

Clearly indicate the attributes of each variable

Indicate the base numbers from which any percentages were computed

Indicate number of cases omitted from table due to missing data

Page 21: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Constructing and Reading Bivariate Tables

Bottom Line for Constructing Tables

– Readers should be able to tell what each variable in the table is, and

– Be able to interpret the overall meaning of this table without having to read the narrative text of the report

Rule of Thumb for Reading Tables

– If table is “percentaged down” then “read across” in making the subgroup comparisons, or

– If table is “percentaged across” then “read down” in making subgroup comparisons

Page 22: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

An Example of a Bivariate Table

Page 23: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Multivariate Analysis

A more complex method that involves analyzing the relationships among several variables

E.g., examining the relationship between an independent and dependent variable while controlling for extraneous variables (recall extraneous, moderating, and control variables in Chapter 7)

Page 24: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

An Example of a Multivariate Table

Page 25: SOWK 6003 Social Work Research Week 10 Quantitative Data Analysis By Dr. Paul Wong

Descriptive Statistics and Qualitative Research

The use of descriptive statistics, which are used to describe characteristics of the sample, often can enrich a qualitative study

It is not uncommon to find quantitative data included in reports of qualitative research – oftentimes counting phenomena is part of detecting patterns in qualitative research