Logistic Regression, Logit Models, and Logistic Discrimination

Abstract

In logistic regression, there is a (binary) response of interest, and predictor variables are used to model the probability of that response. More generally , in a table of counts, primary interest is frequently centered on one factor that constitutes a response (dependent) variable. The other factors in the table are only of interest for their ability to help explain the response variable. Special kinds of models have been developed to handle these situations. In particular, rather than modeling log expected cell counts or log probabilities (as in log-linear models), when there is a response variable, various log odds related to the response variable are modeled.