|
Abstract:
|
The issue of estimation accuracy in multiple regression with small samples has long been a concern. To improve the statistical performance, the bootstrap is employed to obtain more accurate standard errors; however, the bootstrap in regression models has inevitable limitations. The present study applies a new multivariate resampling method, the Sample Smoothing Amplification Resampling Technique (S-SMART), to improve the estimation accuracy in multiple regression with small samples. S-SMART is a distribution-free method and utilizes the kernel smoothing technique to obtain multivariate resamples based on a given small sample while retaining the key statistical properties possessed by the small sample; and therefore, the statistical performance of multiple regression can be improved through S-SMART.
|
