I have been asked to recreate a repeated measures mixed model originally performed in SAS. I have patients on two different treatments, followed-up for 5 visits each. The original analysis assumed a common unstructured covariance matrix among visits for each treatment group. Is this possible in Stata? I've been looking at the example 6 in the multilevel reference manual (http://www.stata.com/manuals13/me.pdf) for inspiration but I'm struggling to understand why the below code doesn't work (on this example data and using my real data) and suspect this reflects a lack of understanding on my part. Any help appreciated.
Code:
webuse exercise.dta, clear gen program1=(program==1) gen program2=(program==2) mixed strength program##day || id: program1, cov(un) || id: program2, cov(un) // Doesn't converge
