Categorical Analysis

Instructor(s): Thomas Pullum, Sociology

"Which candidate did you vote for?" "Which diagnosis is correct given the patient's symptoms?" "Will that offender avoid crime in the future?" Responses to questions of this kind are recorded in unordered categories whose statistical analysis is the topic of this workshop. When observations from individual cases are available, they may be treated within a regression framework in which coefficients for the explanatory variables are found under the assumptions of linear probability, linear discriminant, probit, logit, multinomial logit, or conditional logit models. When individual cases have been grouped into a contingency table, cell proportions rather than individual responses constitute the dependent variable, and linear probability, log-linear or logistic models may be employed. The statistical justification of models for both situations will be presented, and their application to survey data will be illustrated in class and through computer exercises. 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 and Powers and Xie's Statistical Methods for Categorical Data Analysis.