ABSTRACT
This chapter describes the choice for the correlation structure of repeated measures. One must know or have a good idea of the appropriate covariance structure of the repeated measures in order to have the correct standard errors of the estimated coefficients of the mean profile. In the Bayesian context, the correct covariance matrix will have appropriate standard deviations for the posterior distribution of the coefficients involved in the mean profile of the repeated response over time. It should be noted that the covariance matrix of the errors in the mean response depend on the mean response and conversely; thus, the two entities are related. We will take the approach of determining a good regression representation of the mean profile and use that to search for an appropriate covariance or correlation matrix.
