ABSTRACT
This chapter explores the readers to Rindskopf's methods for LR-X2 decomposition. Helmert contrasts are most useful when variable categories can be meaningfully grouped as in this drug example. The grouping leads to a sequence of partitioning steps that are the only ones tested. Groups are formed because members belong together for natural or theoretical reasons. The chapter illustrates and outlines designs for partitioning repeated measures data. For each repeated measures design, the same constraints apply as for log-linear modeling of repeated measures. For instance, one can compare multiple groups in their change patterns; one can analyze several variables in which change may manifest itself, or one may look at more than one observation point. The concepts of division, joining, and decomposition of change patterns are always applicable. The reason for this general applicability is that LR-X2 partitioning can be equivalently expressed in terms of nonstandard log-linear models.
