ICPSR workshop on Categorical Data Analysis: I will be teaching a one-week class on categorical data analysis as part of ICPSR's Summer Program in 2007. The class will be held in Ann Arbor, Michigan from June 11, 2007 to June 15, 2007. The workshop examines the most important regression models for binary, ordinal, nominal and count outcomes. A variety of practical methods for interpreting the nonlinear models are presented. Statistical testing and assessing fit is also illustrated with a series of real-world examples. Click here for more information.

Soc650: Categorical Data Analysis is the second course in sociology’s graduate sequence in applied statistics and is taught Fall Semester. The first course, Soc554, deals with models in which the dependent variable is continuous. These include the linear regression model, seemingly unrelated regressions, and systems of simultaneous equations. Soc650 deals with regression models in which the dependent variable is limited or categorical. Such models include probit, logit, ordered logit, and Poisson regression, among others. For details, click here. The prerequisite for this class is a prior course in regression. Unfortunately, there are more students who want to take Soc650 than there are seats in the class; for enrollment information click here.

Soc651: Models for latent varaibles is taught periodically.

ICPSR workshop on Categorical Data Analysis - the right-hand-side of the model: Binary logit and probit are the most commonly used regression models for categorical outcomes. While these models are similar in many respects to linear regression, interpretation is more complicated since the models are nonlinear. This workshop begins by considering the general objectives of interpretation for any regression-type model and then considers why achieving these objectives is more difficult when models are nonlinear. Building on this framework, the binary regression model is developed as a nonlinear model for predicting the probability of a binary outcome. After a review of methods of estimation, the focus turns to interpretation, specification, and testing. Basic methods of interpretation, including odds ratios, ideal types, predicted probabilities, and graphical methods, are illustrated with real-world examples. We begin by applying these methods to simple models that include only binary and continuous explanatory variables. These models are used to illustrate interpretation, testing, and assessing fit. With the fundamental tools of interpretation and testing in hand, we examine a set of increasingly complex specifications that are often encountered in substantive research. These includes the use of nominal and ordinal explanatory variables, group comparisons, interactions, and additional forms of nonlinearity. While these extensions to model specification are illustrated with binary models, the ideas can be readily adapted to other regression models. Indeed, the binary model is an essential building block for creating models such as ordinal regression, multinomial logit, item response models, bivariate probit, and latent mixture models. These models will be briefly discussed the last day of class. The labs will show how to apply each of the methods discussed in lecture. While no prior knowledge of Stata is assumed, attendees should be familiar with linear regression. Last taught in 2002.