Statistical Methods for the Analysis of Repeated Measurements

Charles S. Davis
Senior Director of Biostatistics, Elan Pharmaceuticals

Saturday, May 3, 2003
8:15-4:15pm
(Note the slightly earlier starting time)
at
California State University, Long Beach
College of Business Administration Building, Room 140

Abstract

Studies in which multiple measurements of a univariate response variable are obtained from each experimental unit are frequently used in biomedical and other types of research. In some cases, repeated measurements are obtained at multiple time points from individual subjects. In other applications, correlated responses may be obtained from experimental units that are families or litters, rather than from individual patients or animals. Although many approaches to the analysis of repeated measurements have been proposed and studied, it is often difficult to select, implement, and apply appropriate statistical methodology.

This workshop surveys traditional and modern methods for the analysis of repeated measurements. The workshop will be presented in four sessions, organized as follows.

1. A brief overview of methods appropriate for one-sample, multi-sample, and regression problems involving repeated measurements of (a) normally distributed outcomes, (b) continuous, non-normal outcomes, and (c) binary, polytomous, and ordered categorical outcomes:

Summary statistic approach;
Classical multivariate methods for the analysis of continuous, normally distributed response variables (MANOVA, profile analysis, growth curve analysis, repeated measures ANOVA);
Methods appropriate for categorical repeated measurements (weighted least squares approach, randomization model methods based on the use of Cochran-Mantel-Haenszel statistics);
Nonparametric methods.

2.    The analysis of repeated measurements using linear mixed models.

3.    The analysis of repeated measurements using extensions of generalized linear model methodology (the Generalized Estimating Equations (GEE) method and its extensions, including extensions to ordered categorical outcome variables).

4.    Questions, examples, and further discussion.

While relevant theory is included in the workshop handouts, the focus will be on the application of appropriate methodologies using readily available software (primarily the SAS MIXED and GENMOD procedures).

More information, including registration forms, schedule, map and directions, and Dr. Davis' biography.