Sankhya: The Indian Journal of Statistics

2000, Volume 62, Series B, Pt. 1, 5--24

A STEP-DOWN PROCEDURE FOR ANALYSIS OF TIME-VARYING COVARIATES IN MULTI-VISIT STUDIES

By

SOLANGE ANDREONI, Laboratórios Pfizer Ltda, Sãao Paula

SUMMARY. In multi-visit studies where several subjects are followed for a certain prespecified period of time, the goal may be viewed as the description of patterns of change of the response variable(s) as a function of covariates, time, and or treatment. Often there are many covariates measured at each visit. In view of the cost involved, one question that may arise is whether it is necessary to have the collection of all covariates at each visit. The objective here is to study the necessity of using the covariates at each time point in the analysis, as opposed to using only the baseline measurements. Some hypotheses of interest are formulated in terms of the associated dispersion matrices adjusted for the covariates, and test statistics for comparisons between them based on the step-down procedure are presented. This study may help to design future investigations, or to stop data collection on covariates early in the study.

AMS (1991) subject classification. Primary 62H05, 62H10, 62H15.

Key words and phrases. longitudinal data, repeated-measures, monotone multiple design multivariate model, likelihood ratio.

Full paper (PDF)

This article in Mathematical Reviews