Telephone: 532-4973     E-mail: goe@ksu.edu

Office hours: T U 11:00-12:00 AM or by apt.

Course Description:

The objective of this course is to provide theoretical knowledge and practical experience in using regression analysis as a quantitative method in social research. Regression analysis is arguably the most commonly used advanced statistical technique in the social sciences. As a key foundation of social research, it is important that students have an understanding of this technique and the ability to use it in their own research. This course will cover material concerning the key theoretical underpinnings of regression analysis as a quantitative method. This will be supplemented with practical experience in using regression analysis by estimating regression models with the SAS statistical software program. It is assumed that students already have substantial knowledge of basic statistics. The key objectives are for students to come away from the course with the ability to use regression analysis in their own research as well as the ability to critically evaluate the use of regression analysis in research conducted by others.

Required Texts:

McClendon, McKee J. 1994. Multiple Regression and Causal Analysis. Itsaca, IL: F.E. Peacock.

Berry, William D. 1993. Understanding Regression Assumptions. Beverly Hills, CA: Sage.

Berry, William D. and Stanley Feldman. 1985. Multiple Regression in Practice. Beverly Hills,  CA: Sage.

Fox, John. 1991. Regression Diagnostics. Beverly Hills, CA: Sage.

Book of Selected Readings. This is available on reserve at Claflin Books & Copies, 1814 Claflin Road.

Recommended Texts:

Spector, Paul E. 1993. SAS Programming for Researchers and Social Scientists. Beverly Hills, CA: Sage.

SAS Institute. 1990. SAS/STAT User's Guide, Volume 1. Version 6, 4th Edition. Cary, NC:SAS.

SAS Institute. 1990. SAS/STAT User's Guide, Volume 2. Version 6, 4th Edition. Cary, NC: SAS

K-State Computing Survival Kit. This provides a guide for using the Unix system at K-State.

The recommended texts provide additional information on using SAS for statistical analysis and using the Unix system to run SAS programs. All of them would be good to have for personal reference.

Grading and Course Requirements:

Grades will be determined by four course requirements: (a) a mid-term exam (30%); (b) a final exam (30%); (c) research assignments (10%); and (d) a final research paper (30%).

Exams -- There will be two exams -- a mid-term and a final. The exams will consist of a combination of problem solving and essay questions on the material covered in the lectures and readings.

Research Assignments -- There will be approximately 7-8 research assignments. All research assignments will be handed out in class. In order to complete the research assignments, it is necessary for each student to obtain a copy of a data set that has been compiled for quantitative analysis. The data set must contain at least 5 variables that could be used to estimate a regression model. One strategy is to ask other professors (including myself) if they have data sets available that you could use for your in-class research. Alternatively, you could compile your own data set for your own research. I would recommend compiling your own (if possible) to gain the experience of doing so. A second requirement is that you also must have an account on the K-State Unix system. The first research assignment will be to set up your data set for analysis by the SAS software program. We will cover SAS basics in class, if necessary, and the Spector text will provide further information. The second research assignment will be to specify both a bivariate and multivariate regression model based on the variables available in the data set that you have obtained. The specification of the dependent variable and independent variables for these models must be based on a theoretical logic that you can defend. In effect, your data set must contain a measure of a dependent variable that can be conceptually viewed as being influenced by the 4 (or more) other independent variables in the data set. The third research assignment will be to conduct univariate analyses of the variables you have selected. The fourth & fifth research assignments will be to estimate the bivariate and multivariate regression models you have specified using the SAS REG procedure. The remaining research assignments will involve using SAS programs to test whether or not the regression models you have estimated meet specific theoretical assumptions of regression analysis. This is important for determining whether your statistical findings can be accepted as being valid and reliable. The written portions of all research assignments must be typed and the assignments turned in on time.

Research Paper -- The final research paper will be a compilation of the in-class research assignments. The paper must describe the research problem addressed by your regression model, present the theory and/or hypotheses that are to be tested by your multivariate regression model, describe how the study variables were operationalized and data collected, provide a univariate analysis of the study variables, present evidence from the tests that you have used to address whether your sample data meet specific theoretical assumptions of regression analysis, describe any data modifications that you used, present the findings from the regression analysis, discuss the substantive meaning of your findings and draw conclusions from your research. All papers must be typed and are due the day of the final exam, December 14.

Policy regarding final papers & incomplete grades: It is my policy that students who complete all their work on time should be rewarded above those who do not. It is strongly encouraged that students turn in the final paper on the day of the final exam (December 14). Under special circumstances, I will give a student an incomplete with the opportunity to receive full credit for the final paper as long as it is turned in by January 18, 2000, five weeks after the final exam. Students who opt to take an incomplete and turn in their papers between January 19, 2000 and May 31, 2000 will automatically receive a 10% reduction on their grade on the paper. Those taking an incomplete and turning in their final papers between June 1,2000 and December 31, 2000 will receive an automatic 20% reduction on their grade on the paper. Those turning their final paper in 2001 or beyond will receive an automatic 30% reduction.

Topic Outline: Readings are listed under each topic section

1.0 Introduction -- Basic Concepts in Causal Analysis

Required Reading: McClendon, Multiple Regression and Causal Analysis, Chapter 1.

2.0 Computer Analysis of Data -- Basics of SAS Programming

Optional Reading: Lecture material on using SAS can be augmented by the Spector  textbook,SAS Programming for Researchers and Social Scientists.

3.0 Using the Unix system at K-State

Optional Reading: K-State Computing Survival Kit

4.0 The Bivariate Regression Model

Required Reading: McClendon, Multiple Regression and Causal Analysis, Chapter 2.
Pedhazur, Chapter 2, Simple Linear Regression & Correlation

5.0 Multiple Regression Analysis

5.1 Estimating Multivariate Regression Models

Required Reading: McClendon, Multiple Regression and Causal Analysis, Chap. 3.
Pedhazur, Chapter 3, Elements of Multiple Regression Analysis

Optional Reading: SAS/STAT User's Guide, Volume 1, Chapter 23, The FREQ  Procedure,
SAS/STAT User's Guide, Volume 2, Chapter 36, The REG  Procedure,

5.2 Sampling Distributions & Tests of Statistical Significance

Required Reading: Mohr, Chapter 3, The Sampling Distribution
Korin, Chapter 10, Sampling Distributions
Mohr, Chapter 5, Significance Testing
Korin, Chapter 12, General Concepts for Tests of Significance
McClendon, Multiple Regression and Causal Analysis, Chapter 4, pp. 133-174.

5.3 Using Binary Variables or "Dummy" Variables as Independent Variables

Required Reading: McClendon, Multiple Regression and Causal Analysis, Chapter 5.
Hardy, Chapter 3, Using Dummy Variables as Regressors

6.0 Digging Deeper: Understanding the Theoretical Assumptions of the Regression Model

Required Reading: Berry & Feldman, Multiple Regression in Practice, Chapter 1, Pp.9-18.
Berry, Understanding Regression Assumptions, Chapters 1-4, Pp. 1-22.
Review McClendon, Multiple Regression and Causal Analysis, Chapter 4,  pp. 142-147.

6.1 The Assumption of No Specification Error -- Modeling Nonlinear & Nonadditive Relationships

Required Reading: Berry, Understanding Regression Assumptions, Pp. 30-41, 60-66.
Berry & Feldman, Multiple Regression in Practice, Pp. 18-26, 51-72.
Fox, Regression Diagnostics, Pp. 1-9, 53-61.
Jaccard, et al., Interaction Effects in Multiple Regression
McClendon, Multiple Regression and Causal Analysis, Chapters 6 & 7.

6.2 The Assumption of No Perfect Collinearity

Required Reading: Berry, Understanding Regression Assumptions, Pp. 24-27.
Berry & Feldman, Multiple Regression in Practice, Pp. 37-50.
Fox, Regression Diagnostics, Pp. 10-21.

6.3 The Assumption of Constant Error Variance or Homoscedasticity and the  Assumption of No Autocorrelation

Required Reading: Berry, Understanding Regression Assumptions, Pp. 67-81.
Berry & Feldman, Multiple Regression in Practice, Pp. 73-89.
Fox, Regression Diagnostics, Pp. 49-53.
McClendon, Multiple Regression and Causal Analysis, Chapter 4,  pp. 174-195.

6.4 The Assumption that the Error Term Must be Normally Distributed

Required Reading: Berry, Understanding Regression Assumptions, Pp. 81-83.
Fox, Regression Diagnostics, Pp. 40-48.

6.5 The Assumption of No Measurement Error

Required Reading: Berry, Understanding Regression Assumptions, Pp. 45-60.
Berry & Feldman, Multiple Regression in Practice, Pp. 26-37.

6.6 The Assumption that Independent Variables Have Non-Zero Variance

6.7 The Assumption that Independent Variables Must Not Be Correlated With the Error Term

Required Reading: Berry, Understanding Regression Assumptions, Pp. 27-29.

6.8 The Assumption that the Expected Value of the Error Term is Zero

Required Reading: Berry, Understanding Regression Assumptions, Pp. 41-45.

7.0 Identifying Influential Cases

Required Reading: Fox, Regression Diagnostics, Pp. 21-40.

Important Dates to Remember

September 28 -- Last day for acquiring a dataset that includes at least 5 variables and 50 cases that can be used for estimating regression models.
October 19 -- Mid term exam (approximately)
December 14 -- Final exam and final paper due