My office is 2146 Watkins Hall, where I have office hours that will be announced in lecture. I may also be trying "virtual office hours" with chat and whiteboard using our course web site. I am also available by appointment. I am frequently in the office, and can be interrupted if my door is open. My office phone number is (909) 787-3638. You may leave phone messages (urgent only) with the Sociology Department at (909) 787-5444. The best way to reach me is email as robert.hanneman@ucr.edu. You can learn more about my interests and work by visiting my home page.
Return to the outline of this page
Sociology 203B is the second course in the basic sequence in applied statistical analysis in the graduate program in Sociology. This course deals primarily with issues of measurement, and how such information can be included in the statistical analysis of relationships among variables. Sociology 203B assumes that students have completed sociology 203A or an equivalent course (contact the instructor if you are unsure about your preparation).
Applying multivariate statistics for prediction and hypothesis testing to the data collected by sociologists involves a number of challenges. The first course in the sequence (203A) deals with the fundamental issues of statistical control and the estimation and interpretation of a number of types of single equation models. Many theories, however, imply models with multiple simultaneous dependent variables. The models examined in 203A assumed that both the dependent and independent variables were directly measured with single, reliable, and valid indicators. Many concepts from sociological theories cannot be measured so readily. Problems with measuring variables seriously limit our ability to use modeling and hypothesis testing.
A major issue area in bridging the gap between sociological theories and observational data is measurement (or, as it is sometimes called "auxiliary theory"). This course will introduce the logic of the classical measurement model and the associated concepts of reliability and validity. A number of methods for modeling variables and constructing indexes and tests will be examined (including principal components, exploratory and confirmatory factor analysis, and latent class analysis). Modeling approaches involving both structural and measurement models (most generally, called "Structural Equation models" or SEMs) will be introduced (frequently called "LISREL" or "COSAN" or "EQS" models after the ML programs used to estimate them).
The measurement models discussed above deal with how we could combine variables, on the basis of observations across cases. But measurement also involves combining cases, on the basis of what we can observe about multiple variables. That is, sociologists also create "Typologies" or "taxonomies" and compare cases (rather than study association among variables) to reach conclusions. We will (briefly) examine some more "qualitative" approaches to measurement that emphasize the scaling of cases and creation of typologies, rather than the scaling of variables to create factors (e.g. Multi-dimensional scaling, cluster analysis, Qualitative Comparative Analysis).
Return to the outline of this page
Meetings: The course will meet for lecture each Thursday of the Spring Quarter from
4:10 to 7:00 p.m. in Watkins 2140. From time to time, we may also use the computer lab.
Resources: Students are expected to read their e-mail, discussion forums, and the calendar for course events posted
in the web site. Homework assignments will require the use of SAS and EQS. Both are available on the
wizard computer. If you do not have an account, Mr. Shoon Lio can provide you with one, and show you how
to use the system.
Reading materials for Sociology 203B will be:
Larry Hatcher. 1994. A Step-by-Step Approach to Using the SAS System for Factor Analysis and Structural
Equation Modeling. Cary, N.C.: SAS Institute, Inc.
Peter M. Bentler. 1995. EQS: Structural Equations Program Manual. Encino, CA: Multivariate
Software, Inc.
Robert A. Hanneman. Annotated examples of analyses.
Various. Photocopied readings cited below.
Requirements and grades: Students will be graded on the basis of a mid-term examination (given after we complete
the multiple-equation modeling material), worth 25% of the grade, and a series of homework assignments that are
described below. Papers must be submitted at the times announced in lecture unless the instructor grants specific
prior exceptions.
Return to the outline
of this page
|
Topic |
Readings |
| Introduction: Measurement, Latent Variables and Latent Types | |
| Latent Variables | |
| Classical test theory, reliability and validity assessment | Carmines and Zeller (Xerox), all. Hatcher, Chapter 3 exercise on reliability and coefficient alpha |
| Principal components analysis | Hatcher, Chapter 1 |
| Exploratory factor analysis | Hatcher, Chapter 2 |
| Structural equation models in observed variables ("Path" analysis) | Hatcher, Chapter 4; Bentler, p.13-25 |
| Confirmatory factor analysis | Hatcher, Chapter 5; Bentler, p. 26-32 |
| Structural equation models with latent variables | Hatcher, Chapter 6; Bentler, p. 33-38 web reading on identification of structural equation models |
| Multiple group comparisons with SEM | Bentler, Chapter 7 |
| Latent Types | |
| "Qualitative Scaling" or Classification | Bailey (Xerox), chapters 1, 2. |
| Cluster Analysis and related techniques | Bailey (Xerox), chapters 3, 4, 5. |
Return to the outline of this page