An Information Analysis on Modeling Interaction Effects in Logistic Regression

The Akaike information criterion (AIC) is commonly used to select a logistic regression model for optimal prediction of a binary response by a specified family of models. It however lacks a convincing method of prescribing a proper family of models using the desired predictors and their interaction effects. For an alternative approach to model selection, we propose a direct selection scheme which first identifies the indispensable regressors as main-effect predictors, then examines significant interaction effects between the selected predictors such that a logistic model is constructed. The two-step selection scheme is formulated by testing for valid information identity between the response and the predictors, from which the most parsimonious logistic model is derived from the least set of indispensable predictors and interaction effects. As a byproduct, the minimum AIC model is easily found in a neighborhood of the selected model. The scheme is employed to yield the logistic model for predicting the acquisition of professional licenses in a survey of employed youth workers.
Comment: Following the JASA article (2010) by Cheng et al., this article introduces a constructive scheme for selecting parsimonious models for a multivariate contingency table using logistic regression (i.e., a well-defined target variable plus several predictors). Please email your comments and questions to the corresponding author