Variations in choice sets and identification of mixed logit models: Monte Carlo evidence
2008
Lecocq, Sebastien
Empirical identification of Mixed Logit models is known to depend on the richness of the data in terms of variations in the explanatory variables. In this paper, we wonder whether choice set variations observed in scanner-based consumer surveys are sufficient to enable identification. Using Monte Carlo experiments, we show that when a random parameter is applied to a binary variable, Mixed Logit models are identified only if the number of options varies across a sufficient number of individuals and/or choice situations. Conversely, when a random parameter is applied to a continuous variable, models are identified without any choice set variation.
اظهر المزيد [+] اقل [-]