Categorical Data Analysis


  • Shawna Smith, University of Michigan

This class focuses on the basic regression models for categorical dependent variables. While advances in software have made it simple to estimate these models, post-estimation interpretation is difficult due to the nonlinearities of the models. The class begins by considering the general objectives for interpreting the results of any regression type model and then considers why achieving these objectives is more difficult with nonlinear models. Basic concepts and notation are introduced through a review of the linear regression model. Within this familiar context, the method of maximum likelihood estimation is presented. These ideas are used to develop the logit and probit models for binary outcomes. A variety of practical methods for interpreting nonlinear models are presented. The models and methods of interpretation for binary outcomes are extended to ordinal outcomes using the ordinal logit and probit models. The multinomial logit model for nominal outcomes is then discussed. Finally, a series of models for count data, including Poisson regression, negative binomial regression, and zero modified models are presented. A major component of the course is using Stata to estimate and interpret the models and particularly the special commands for post-estimation interpretation. The course assumes familiarity with the linear regression model. Familiarity with Stata is not assumed.

Participants should enter this workshop with an active working knowledge of the topics covered in Regression Analysis II: Linear Models and Mathematics for Social Scientists, II. Readings will be drawn from texts such as Long's Regression Models for Categorical and Limited Dependent Variables.

Fees: Consult the fee structure.

Tags: logit, probit, nonlinear

Course Sections

Section 1

Location: ICPSR -- Ann Arbor, MI

Date(s): July 24 - August 18

Time: 3:00 PM - 5:00 PM


  • Shawna Smith, University of Michigan