A1272
Title: A cross-validation bandwidth choice for nonparametric tests in regression models
Authors: Sandie Ferrigno - INRIA (France) [presenting]
Marie-Jose Martinez - University of Grenoble (France)
Abstract: Many goodness-of-fit tests assess the different assumptions of a regression model. The focus is on the task of choosing the structural part of the variance function in the regression model, and three nonparametric tests are considered, all based on generalizations of the Cramer-von Mises statistic. Nonparametric kernel methods are used for regression and variance function estimations. Bandwidth selection is of key practical importance for these estimations. The bandwidths for the regression and variance functions are obtained separately by cross-validation methods. A simulation study, based on wild bootstrap methods, is carried out to compare the three tests in terms of statistical significance and power function.
