Abstract

The jackknife method, a resampling technique, has been widely used for statistical tests for years. The pseudo value based jackknife method (defined as pseudo jackknife method) is commonly used to reduce the bias for an estimate; however, sometimes it could result in large variation for an estimate and thus reduce the power for parameters of interest. In this study, a non-pseudo value based jackknife method (defined as non-pseudo jackknife method) was used for testing variance components under mixed linear models. We compared this non-pseudo value based jackknife method and the pseudo value based method by simulation regarding their biases, Type I errors, and powers. Our simulated results showed that biases obtained by the two jackknife methods are very similar; however, the non-pseudo value based method had higher testing powers than the pseudo value based method while the non-pseudo value based method had lower Type I error rates than the preset nomial probability values. Thus, we concluded that the non-pseudo value based jackknife method is superior to the pseudo value based method for testing variance components under a general mixed linear model.

Creative Commons License


This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Share

COinS
 
Apr 27th, 8:30 AM

TESTING VARIANCE COMPONENTS BY TWO JACKKNIFE METHODS

The jackknife method, a resampling technique, has been widely used for statistical tests for years. The pseudo value based jackknife method (defined as pseudo jackknife method) is commonly used to reduce the bias for an estimate; however, sometimes it could result in large variation for an estimate and thus reduce the power for parameters of interest. In this study, a non-pseudo value based jackknife method (defined as non-pseudo jackknife method) was used for testing variance components under mixed linear models. We compared this non-pseudo value based jackknife method and the pseudo value based method by simulation regarding their biases, Type I errors, and powers. Our simulated results showed that biases obtained by the two jackknife methods are very similar; however, the non-pseudo value based method had higher testing powers than the pseudo value based method while the non-pseudo value based method had lower Type I error rates than the preset nomial probability values. Thus, we concluded that the non-pseudo value based jackknife method is superior to the pseudo value based method for testing variance components under a general mixed linear model.