A0673
Title: Heteroscedasticity tests for nonlinear regression models
Authors: Xu Guo - Beijing Normal University (China) [presenting]
Abstract: An efficient test statistic is developed for heteroscedasticity check for nonlinear regression model. Our proposed test statistic can avoid the so called curse of dimensionality. The proposed method requires no bandwidth selection, is simple to compute, based merely on pairwise distances between points in the sample. Asymptotic results are shown. It is proven that our proposed test can converge to finite limit at the rate of $1/n$ under the null hypothesis and can detect any local alternatives which converge to the null hypothesis at the rate of $1/\sqrt{n}$. These results are surprising and interesting. Asymptotically, the dimension of the covariates has no effect on the convergence rate. Some simulation studies are conducted to illustrate the proposed test statistic.
