Testing for superiority between two variance functions

PARDO FERNANDEZ, JUAN CARLOS; BOENTE, GRACIELA

Londres

Conferencia; XII International Conference of the ERCIM Working Group on Computing & Statistics; 2019

Birbek University of London

The focus is on the problem of testing the null hypothesis that the variance functions of two populations are equal versus one-sided alternativesunder a general nonparametric heteroscedastic regression model. The asymptotic distribution of the test statistic is studied under the null hypothesisand under root-n contiguous alternatives. A Monte Carlo study is performed to analyse the finite sample behaviour of the proposed test.