Binary Logistic Regression

This chapter addresses binary logistic regression, a procedure used to analyze the effects of categorical and continuous explanatory measures on a dichotomous response variable. In doing so, it uses examples based on the horse-racing data as well as the 2008 American National Election Study and the 2012 Monitoring the Future study. In addition to binary logistic regression, communication scholars can apply multinomial and ordinal logistic regression, both of which are covered. The chapter also reviews the fundamental components of binary logistic regression, comparing and contrasting it with ordinary least squares (OLS) regression. OLS models are based on least squares estimation techniques, while estimates in the logistic regression model are based on maximum likelihood. The chapter further outlines parameter estimates and standard errors, Wald statistics, log-likelihood measures, and values for model chi-squared. It explores Categorical Variable Codings that show both frequencies and internal coding for explanatory measures.