05-03-2016 02:16 AM
I have a dataset with a continuous outcome variable and 3 treatments (1 of which is placebo). I have about 60 observations in my dataset; 60 subjects each of which got 1 treatment and was measured once - the dataset is balanced but measurements were taken in blocks of days but only 2 measurements were taken each day (incomplete blocks?). I need to include the block in my model as a random effect. I am trying to get the effect of the drug and im running a proc mixed with a random block effect, but i am not getting an anova table and am getting an error that the "Estimated G matrix is not positive definite." Can anyone tell me what I am doing wrong? Thank you.
05-05-2016 01:54 PM
Please share your PROC MIXED code and enough of your data to enable some trouble shooting. I suspect that you may have misspecified the random effect, as the message 'Estimated G matrix is not positive definite' means that at least one variance parameter is being set to zero. There are a couple of ways around this, but it depends on design, data structure and code.