Logistic Regression (Logit) and Probit Models

This chapter discusses logistic regression and probit models. When the outcome variable is categorical and dichotomous, binary logistic and binary probit models are used for classification, given the predictors. We discuss the concept of logistic regression and the concepts associated with the logit model. We also discuss the approaches to model comparison and selection using analysis of variance. Then we explain the concept of pseudo R-squared and its use in model comparison and selection. We then generate the logistic curve to get a visual representation of the classification, given the predictors. We then discuss the probit model, its assumptions and explain the situations in which the probit model should be preferred to the logit model and vice versa.