July 16th 2015
Faculty of Economic and Business Sciences | Salón de Grados
A nonparametric Bayesian test of fit
2015/07/16 – 13:00 h | María J. Polidoro, Instituto Politécnico do Porto, ESTGF/CIICESI & CEAUL
Abstract
The basis for several statistical methodologies assumes that a specified parametric probabilistic model fits a observed data set. If this assumption does not hold, the quality of the inferences is doubtful. One of the solutions proposed by the Bayesian approach, to study the adequacy of a model, is to define a Bayesian nonparametric alternative model that embed the parametric model under study. Next, to study the adequacy of the model, we use measures of comparison between the two models. The Bayes factor is one of the most relevant of such measures.
In this work, we propose a nonparametric Bayesian test of fit to study the adequacy of the exponential model, using as Bayesian alternative model a mixture of Pólya trees. It is also referred some practical examples and the performance of the Bayesian test of fit is compared, through a simulation study, with some of the classic tests.
Gallery
