ABSTRACT
The logistic regression model describes the relationship between a discrete outcome variable, the “response,” and a set of explanatory variables. A fitted model provides both statistical inference and prediction, accompanied by measures of uncertainty. Data visualization methods for discrete response data must often rely on smoothing techniques, including both direct, non-parametric smoothing and the implicit smoothing that results from a fitted parametric model. For a model with two or more explanatory variables, full-model plots display the fitted response surface for all predictors together, rather than stratified by conditioning variables. The essence is to calculate fitted values for the model terms involving these variables and all low-order relatives, as these variables are allowed to vary over their range. For logistic regression models, they solve the problem of the trade-off between plots on the logit scale, which have a simple representation in terms of additive effects, and plots on the probability scale which are usually simpler to understand.
