A0153
Title: Bayesian analysis for heterogeneity
Authors: Jaeyoung Kim - Seoul National University (Korea, South) [presenting]
Jaebeom Ahn - Seoul National University (Korea, South)
Abstract: A convenient new approach is studied for analyzing empirical problems in the presence of heterogeneity. The approach is a Bayesian semi-parametric method with an empirical prior capturing heterogeneity. The considered heterogeneity is a general type of heterogeneity that covers a variety of cases studied in economic theory and econometric practice. It is also applicable for developing inference for model misspecification. To develop our framework to get empirical prior, we adopt a minimum distance method with an information distance to solve an ill-posed deconvolution problem. Inferential issues of detection and estimation of heterogeneity are also studied based on the framework. The methods are evaluated by a pair of MonteCarlo experiments whose results confirm usefulness of our methods for practical applications. We also applied our methods for two interesting empirical examples.
