Iclogit: a Stata module for estimating a mixed logit model with discrete mixing distribution via the Expectation-Maximization algorithm

This paper describe Iclogit, a Stata module to fit latent class logit models through the Expectation-Maximization algorithm. The stability of this estimation method allows overcoming some of the computational difficulties that normally arise when fitting such models with many latent classes. This, in turn, permits users to estimate nonparameterically the mixing distribution of the random coefficients because the more the mass points of the latent class model, the better the approximation of the unknown joint density of the random coefficients.