Tuesday, October 4, 2022 at 11:15am to 12:45pm
Uris Hall, 498
Yuichi Kitamura, Yale University
Group Membership in Flexible Choice Models
Abstract: This talk presents various approaches for controlling for unknown membership in choice models. The topic is relevant for nonparametric identification of choice models with unobserved heterogeneity, but also has important implications for clustering algorithms. First, we briefly review approaches for nonparametric identification of mixture models for continuous choice and nonparametric treatment of correlated random effects. Second, we discuss new results on nonparametric identification of discrete choice models with unknown group membership. The main result concerns binary choice models, which carry less information about group membership relative to continuous choice models and thereby calling for more intricate treatments. We discuss nonparametric identification and approaches for "nonparametric model-based clustering." These results directly extend to multinomial choice models. If time permits, related topics such as nonparametric analysis of network models, in particular the Stochastic Block Model (SBM), are discussed.