Permutation Test for Comparing Zero-Inflated Continuous Distributions

• Activity Number: 341
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Government Statistics
• Author(s): Jixiang Wu and Lei Zhang+ and William Johnson*+
• Companies: South Dakota State University and Mississippi State Department of Health and Pennington Biomedical Research Center
• Keywords: Permutation ; Zero-inflated ; likelihood ratio ; Wald test ; central limit theorem

Abstract

In this study, the permutation test was used to control the empirical Type I errors and testing powers when using several methods to compare population means when the random variables of interest have zero-inflated exponential distributions. The simulation results indicated that the permutation test can be used to adjust the type I errors for small data sets. Four tests including likelihood ratio, Wald test, central limit theorem, and modified central limit theorem, had the similar type I errors and testing powers for large sample size with and without using the permutation test. Type I errors and test powers can be adjusted by using the permutation test even when inappropriate assumptions are made. Therefore, the permutation test should be a valuable addition to the current statistical methods for comparing population means with underlying zero-inflated data structures.