An Application of a New Multivariate Resampling Method to Multiple Regression

Abstract

This study presents a new multivariate resampling method to improve the performance of multiple regression with small samples. The kernel resamping technique (KRT) is utilized in the multivariate resampling procedure to draw random resamples with random noises, which facilitates obtaining more accurate parameter estimates and their standard errors in multiple regression. The findings from an empirical example suggest that the statistical performance of multiple regression is improved through the KRT technique.