One method of using repeated measures data to compare treatment groups
in a clinical trial is to summarize each subject's outcomes with a si
ngle summary statistic, and then perform a distribution-free compariso
n based on the resulting statistics. We examine extensions of this app
roach and conditions under which they retain proper size in the presen
ce of missing data. The asymptotic relative efficiencies of several su
mmary statistic tests are calculated to show which perform best in a v
ariety of situations. The techniques are illustrated using data from a
n AIDS clinical trial.