Evaluating the Power of Testing Mean Differences Using Glimmix and Mix Procedures

• Activity Number: 397
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Section for Statistical Programmers and Analysts
• Author(s): Achut Adhikari*+ and Robert Steiner
• Companies: University of Northern Colorado and New Mexico State University
• Keywords: Fixed effects ; Random effects ; Mixed models ; Generalized linear mixed models ; Nested ; Power

Abstract

Evaluating the Power of Testing Mean Differences Using Glimmix and Mix Procedures Achut P Adhikari and Robert Steiner University of Northern Colorado and New Mexico State University Abstract In the presence of non-normal response distribution for testing the mean differences in an Analysis of Variance (ANOVA) setting, this study compared the powers of two procedures (Proc Glimmix and Proc Mix) using the widely known Statistical Analysis System (SAS). Data simulation techniques were used to mimic 8 families from Anglo and 8 families from Hispanic ethnicity. Each family was measured three times. The families were nested within ethnic groups. The three fixed effects (ethnicity, time and interaction between ethnicity and time) were tested and their powers were compared using these two SAS procedures. In two of the three hypotheses testing, Proc Glimmix was found to have higher power to Proc mix procedure while the powers were found identical in one hypothesis testing. It appeared that Proc Glimmix performed at least equal or better than Proc mix procedure when the response distribution is skewed. Key words: fixed effects, random effects, mixed models, generalized linear mixed models,