Covariance Structures for Multiple Repeated Measures Models

• Activity Number: 498
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Hongmei Han*+ and Robbie Beyl and Lei Zhang and William Johnson and Jeff Burton
• Companies: Pennington Biomedical Research Center and Pennington Biomedical Research Center and Mississippi State Department of Health and Pennington Biomedical Research Center and Pennington Biomedical Research Center
• Keywords: Covariance patterns ; General linear mixed models ; Longitudinal data

Abstract

Consider a statistical model with a single outcome that is observed repeatedly under different circumstances. In a study employing an oral glucose tolerance test, for example, serum glucose levels may be measured just prior to ingesting a dose of glucose and then repeatedly every 30 minutes for a total of 120 minutes. The aim is to determine the pattern of change in glucose levels as an indication of how efficiently the individual is disposing the glucose from the blood. If an intervention is given and the process is repeated under this new condition, the data may be analyzed using a doubly repeated measures model. The use of structured patterns in the underlying covariance matrix for correlated residuals in statistical models involving repeated measures is well documented. In models involving multiply repeated measures, however, use of different covariance patterns for different conditions has not been fully discussed. The purpose of this paper is to fill that need by presenting an illustrative overview in terms of practical examples.