Bayesian Local Influence of Semiparametric Structural Equation Models

• Activity Number: 535
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Ming Ouyang* and Xiaodong Yan and Niansheng Tang and Xinyuan Song
• Companies: The Chinese University of Hong Kong and Yunnan University and Yunnan University and
• Keywords: Bayesian local influence ; perturbation schemes ; latent variables ; semiparametric modeling ; MCMC method

Abstract

This research develops a Bayesian local influence method for semiparametric structural equation models. The effects of minor perturbations to individual observations, sampling distributions, and prior distributions on the statistical inference are assessed with the use of various perturbation schemes. We construct a Bayesian perturbation manifold to characterize such perturbation schemes.The first- and second-order influence measures are proposed to quantify the degree of minor perturbations to different aspects of a statistical model on the basis of a variety of objective functions such as Bayes factor, \phi-divergence, and posterior mean distance. We conduct simulation studies to evaluate the empirical performance of the Bayesian local influence procedure and illustrate the proposed methodology via a real application.