Online Program Home
  My Program

Abstract Details

Activity Number: 129 - Quantile and Nonparametric Regression Models
Type: Contributed
Date/Time: Monday, July 31, 2017 : 8:30 AM to 10:20 AM
Sponsor: Section on Nonparametric Statistics
Abstract #323640
Title: Learning Non-Parametric Binary Regression Using Flixible Power Link Function with GP Priors
Author(s): Abhishek Bishoyi* and Xiaojing Wang and Dipak K Dey
Companies: University of Connecticut and University of Connecticut and university of connecticut
Keywords: Gaussian process ; Nonparametric regression ; Binary measurements ; Flexible link ; Bayesian computation ; Surrogate slice sampling
Abstract:

Binary measurements are commonly used in many scientific fields. There are two critical issue in Binary regression problem. First critical issue in modeling binary response data is the appropriate choice of link functions. Commonly used link functions such as logit, probit or c-loglog have fixed skewness and lack flexibility to allow data to determine the degree of skewness. Second, in many cases simple monotonic trend may not be appropriate for the latent regression function. To address both these limitation, we propose a new family of flexible binary regression model which combines power link function with a Gaussian process prior on the latent structure. Model parameters are estimated using Bayesian computation. The performance of the proposed model are illustrated through detailed simulation studies and a real data example.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

 
 
Copyright © American Statistical Association