Conditioning in the growth curve model

We consider a priori reduction of the number of conditioning variables or covariates in the growth curve model. Such reduction depends on the relation between the inverse covariance matrix and the profile design matrix. If the covariance matrix arises from an autoregressive process and the profile design matrix belongs to a certain class, the number of covariates required may be reduced. The extent of the reduction may be examined in a systematic fashion and depends on the order of the process. The random coefficients model is discussed briefly and an example is presented.