Bayesian Analysis of Nested Logit Model by Markov Chain Monte Carlo

We develop a Markov Chain Monte Carlo algorithm for estimating nested logit models in a Bayesian framework. Appropriate "heating target" and reparametrization techniques are adopted for fast mixing. For illustrative purposes, we have implemented the algorithm on two real-life examples involving 3-level structures. The first example involves Social Security's disability determination process (Soc. Security Bull. 58 (1995)). The second one is taken from Amemiya and Shimono's (Econ. Stud. Q. 40 (1989)) model of labor supply behavior of the aged. We applied a combination of various convergence criteria to ensure that the chain has converged to its target distribution