Hypothesis Testing Using Small Samples of Repeated Measures Data

• Activity Number: 422
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Xueliang Pan*+ and Xiaobai Li and David Jarjoura
• Companies: and The Ohio State University and Ohio State University
• Keywords: hypothesis testing ; repeated measures ; small sample ; mixed model ; random contrast ; convergence

Abstract

A typical ex-vivo experiment with a simple 2x2 design uses a small number of samples and each sample placed under all experimental conditions multiple times. To test the interaction, mixed models with selected covariance structure and degrees of freedom are used. However, with small sample sizes and potential missing data, the mixed model with desired covariance structure may fail to converge. Overly simplified covariance structures are assumed instead, and this can inflate type I error. We propose a t-test approach and a random contrast approach without selecting covariance structures. Theoretically we show how these approaches work in balanced data. Simulations show that both methods are robust to model assumptions while controlling type I error and achieving power appropriately. Furthermore, convergence issues are rare for our methods in the presence of various missing data patterns.