Estimation and Inference for Conditional Distribution Models with Response-Dependent Index Coefficients

• Activity Number: 41
• Type: Contributed
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Ming-Yueh Huang*+
• Companies: National Taiwan University
• Keywords: conditional distribution model ; semiparametric ; varying-coefficient ; Gaussian process ; single-index ; uniform consistency

Abstract

To characterize the conditional distribution of a real-valued response conditioning on a set of explanatory variables, we propose a more flexible semiparametric regression model, which allows a part of index coefficients varying with response values. The presented model generalizes most of classical single-index conditional distribution models and is able to accommodate variant types of data structures. Based on the induced response process, an easily implemented pseudo likelihood estimation approach is developed to estimate the response-varying/invariant coefficients. Obviously, as shown in the literature, the resulting parameter estimator at each response level can be shown to be consistent, asymptotically normal, and semiparametrically efficient. In this study, we further established a general theoretical framework for the uniform consistency and limiting Gaussian process of the proposed functional estimator.