Generalized Maximum Entropy Estimation of Discrete Choice Models

Abstract

In this article, we describe the gmentropylogit command, which implements the generalized maximum entropy estimation methodology for discrete choice models. This information theoretic procedure is preferred over its maximum likelihood counterparts because it is more efficient, avoids strong parametric assumptions, works well when the sample size is small, performs well when the covariates are highly correlated, and functions well when the matrix is ill conditioned. Here we introduce the generalized maximum entropy procedure and provide an example using the gmentropylogit command.