Power Analysis of a Left-Truncated Normal Mixture Distribution with Applications in Red Blood Cell Velocities

• Activity Number: 82
• Type: Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Huichao Chen*+ and Xiangjin Xu
• Companies: Harvard School of Public Health and Binghamton University
• Keywords: mixture distribution ; mixing proportion ; LRT

Abstract

In mice infected with malaria, the splenic red blood cell (RBC) velocity profile suggests a left-truncated normal mixture distribution with two components. Since regularity conditions do not hold for the likelihood ratio test (LRT) statistic to have its usual asymptotic null distribution of chi-squared, we study the sample size requirements of such mixture models by first conducting Monte Carlo simulations to estimate the percentiles of the LRT statistic. For mixing proportions ranging from 0 to 1, we then investigate how large the sample size would have to be to detect the second component with a reasonable power, given a setting reflected by the splenic RBC velocities.