If you are looking to fit a random slopes model (in har_y), this looks good. I still think of it as a random intercepts model, with a single G side random statement: random intercept cult blk/subject=loc type=un; I end up thinking this because of the R side factors including all of these variance components in an unstructured covariance structure matrix. My experience would be that the output would throw a message that the G matrix was not positive definite--basically saying that the R part was explaining all of the variability. However, with sufficient data, it could be done. Steve Denham
... View more
