Statistical Engineering Division SeminarReview and Evaluation of Bayesian Diagnostic Techniques for Detecting Hierarchical Structure
Guofen Yan Abstract Motivated by an increasing number of Bayesian hierarchical model applications, we evaluate the performance of several diagnostic techniques when the fitted model includes some hierarchical structure, but the data are from a model with additional, unknown hierarchical structure. Our investigation suggests two promising techniques: (a) distribution of the set of individual posterior predictive p values, (b) the conventional posterior predictive p value with the F statistic as a checking function. We will start with a review of Bayesian model diagnostic approach, highlighting posterior predictive approach. We will present the numerical examples and a real data example to illustrate the methods and evaluation. NIST Contact: Charles Hagwood, (301) 975-2846.
Date created: 8/20/2007 |