Bayesian and classical estimation of mixed logit model for simulated experimental data

This paper explores the similarities and differences between classical maximum Simulated Likelihood and Bayesian methods in estimating parameters of mixed logit model. We use the simulated dataset to numerically evaluate the performance of the two methods. Our results show that two methods provide close estimates in our study; both methods are fairly straightforward to implement. We also find that classical approach is faster than Bayesian method and Bayesian method can be sensitive to given parameters. The results suggest that the choice between the two estimation approaches depends more on researcher's preference.