Gunnar Almgren

Winter, 2002

(206) 685-4077 mukboy@u.washington.edu

Office Hours: Tuesday 12:30-1:30 PM

Class Times: 9:30-10:50 T,Th -SSW 125

Chan-woo Kim

chanwoo@u.washington.edu

Social Welfare 588 is the second quarter of a two-quarter sequence in statistical methods for social welfare research required for first year Ph.D. students. The basic objective of Social Welfare 588 is to establish a foundation multivariate methods based on linear models. Methods covered include factorial analysis of variance, multiple regression, and analysis of covariance. The course will consider the ways in which these methods can be employed to add to the knowledge base in social welfare, as well as their particular limitatations and common pitfalls.

General Course Objectives:

Upon completing the course, students will be able to:

1. Understand essential aspects of the relationship between statistical analysis, research design, and measurement as applied to social welfare research.

2. Apply, compute, and interpret advanced descriptive statistics as called for by the nature of the data and the essential questions.

3. Apply, compute, and interpret inferential statistics to test hypotheses from experimental designs (including factorial designs).

4. Apply, compute, and interpret inferential statistics from survey research designs.

5. Test for interaction effects in multivariate analysis.

6. Understand and appropriately apply statistical controls as indicated by the essential characteristics of the research design and relevant questions.

7. Be able to critically evaluate the use of advanced descriptive and inferential statistics for experimental and survey design in the published research.

8. Be able to use SPSS for Windows for data management and analysis.

9. Understand the ways in which statistics can be misused in social welfare research to distort truth and perpetuate injustice.

10. Articulate a philosophy of honest and ethical use of statistics and the reporting of research results.

Bohrnstedt and Knoke. Statistics for Social Data Analysis. 3^rd Edition. Itaska, IL: Peacock Publishers.

Pedhazur. Multiple Regression in Behavioral Research. 3^rd Edition. New York: Harcourt-Brace.

Allison. Multiple Regression: A Primer. 1999 Pine Forge Press.

Jaccard, Turrisi and Wan. Interaction Effects in Multiple Regression. Sage Series on Quantitative Applications in the Social Sciences No. 72. 1990 Sage Publications.

Learning will be based on assigned readings, lecture/discussion sessions, weekly computer lab assignments and preparation for two exams taken over the course of the quarter. Lectures/discussions sessions will be held twice per week, which will follow readings assigned from the main course text, as well as occasional readings assigned from course reserves.

Lab assignments will be 20% of the course grade, 50% of the grade will be based on an average of two mid-terms exams, and 30% of the course grade will be based on a project chosen by the student that employs multivariate analysis to respond to a research question.

Exam 1 January 29^th

Exam 2 February 26^th

Project Due Date March 12^th

Session 1 January 8

Factors Affecting the Precision of the Regression Equation

Interpreting SPSS Output in a Bivariate Regression Equation

Basic Assumptions and their Typical Violations

Session 2 January 10

Elements of Multiple Regression: Two Independent Variable Models

The OLS Multiple Regression Equation: Basic Notation and Interpretation

Calculation of basic statistics using sums of squares

The Standard Score Regression Equation

Comparing the interpretation of the slope and standardized regression coefficient

Calculation and Interpretation of the squared multiple correlation (R-squared)

Readings

Text: Chapter 6

Session 1 January 15

Tests of Significance and their Interpretation in Multiple Regression

Basic Regression Diagnostics

Session 2 January 17

Statistical Controls: Comparison with other Approaches to Control

Partial and Semipartial Correlation

Readings:

Text: Chapter 7and 8

Session 1 January 22

Prediction vs. Explanation

Approaches to Predictor Selection and Elimination

Session 2 January 24

Explanation: Basic Approaches and Major Sources of Error

Specification Errors, Measurement Errors, and Collinearity

Readings

Pedhazur, Chapter 8

Pedhazur, Chapter 10

Session 1 January 29

Session 2 January 31

Nominal Independents: Dummy Coding

Nominal Independents: Alternative Approaches with Multiple Categories

Readings:

Pedhazur, Chapter 11

Session 1 Feb. 5

Accounting for Interaction Effects

Interpreting Main and Interaction Effects in the Presence of Interaction

Session 2 February 7

Curvilinear Models

Readings:

Borhnsedt and Knoke (3^rd), Statistics for Social Data Analysis, pp. 309-311

Jaccard, Turrisi and Wan Sage Series No. 72, pp. 7-31

Pedhazur, Chapter 12, 495-501

Pehazur, Chapter 13: pp. 513-547

Session 1 February 12

Alternative Approaches to Coding Categorical Variables

Regression with Two Categorical Variables

Partitioning the Regression Sum of Squares

Session 2 February 14

Applying MANOVA to Analysis of Simple Effects

Reading:

Pedhazur, Chapter 12: pp. 410-441

Session 1 February 19

The Logic of ANCOVA

Alternative Applications of ANCOVA

Session 2 February 21

Logistic Transformation and the Estimation of Dichotomous Outcome Models

Estimation of Parameters and Model Fit

Readings:

Text: pp. 303-309 and Chapter 9

Session 1 February 26

Session 2 February 28

Causal Models and Path Analysis

Text: Chapter 11

Session 1 March 5

A Basic Overview of Structural Equation with Latent Variables

Session 2 March 7

A Basic Overview of PC Factor Analysis

Readings: TBA

Session 1 March 12

Analysis Project Reports and Critiques

Session 2 March 14

Analysis Project Reports and Critiques

Course Evaluation