05-08-2013 06:12 PM
I have a large longitudinal dataset, with around 750,000 observations and 200,000 individuals. I used proc hpmixed to analyze the data. I noticed that proc hpmixed do not have the option to get covariance matrix of fixed effect parameters. It is "covb" in proc mixed procedure. I tried to pass the covariance results of random effects from proc hpmixed to proc mixed, and hold the values in parm and option noiter in proc mixed, but proc mixed still cannot handle such big data due to insufficient memory . Is there other way to get covariance matrix of fixed effect parameters?
05-10-2013 01:13 PM
Some more info might help. Can you share both your HPMIXED and MIXED code, and tell us something about the variables? Anything I can come up with will depend on the variables and their relationships.