A Bayesian analysis of tree structure specification in nested logit models

This paper adopts a Bayesian approach to the problem of tree structure specification in nested logit models. I use the Laplace approximation and Reversible Jump Markov Chain Monte Carlo (RJMCMC) to estimate marginal likelihoods in both a simulated and a travel mode choice data set. I find that the Laplace approximation is remarkably accurate, and that model selection is invariant to prior specification.