I am interested in constructing an A-optimal design for a Bayesian linear model with a diagonal error covariance matrix where the error variance depends on one of the categorical variables in the analysis. I was wondering if and how SAS proc optex could be used to construct an optimal design for this problem?