04-09-2013 09:31 PM
PROC GLIMMIX DATA=blah method = quad;
CLASS sub team trial;
MODEL growth = team ;
RANDOM trial / SUBJECT=sub residual TYPE=CS;
RANDOM trial trial*team;
I get an error stating that QUAD does not work for R-side effects.
04-10-2013 08:34 AM
Thaink about what you need here--since growth is assumed to be normally distributed (at least to have Gaussian error), you probably don't need method=quad. The REML estimates should give what you are looking for. However, if you do want to include the residual error in the optimization, then you could remove the residual option from the first RANDOM statement.