I agree with you Steve. For the ddf concerns, I would add the ddfm=kr option in the MODEL statement. ... View more

What is in this Covariance Parameter Estimates table? Can you send it here? ... View more

Did you get the Covariance Parameter Estimates table in the output? ... View more

There are several ways to model the correlations in your data. I am not sure what the variables GroupA, GroupB, and GroupC are, but one of the straightforward models would be modeling the correlations through random effects. For example,   PROC MIXED; CLASS person groupA groupB groupC; MODEL DV = groupA|groupB|groupC; RANDOM intercept GroupA GroupB GroupC GroupA*GroupB GroupA*GroupC GroupB*GroupC/subject=person; run; Some of the random interaction effects might or might be present in your data. If you get zero estimated variance for some of the random effects, you might consider taking them out of the RANDOM statement. ... View more

Actually, as far as I know, there is no consensus on what to do when the G matrix is not positive definite. While some advice removing the "offending" random effects, others argue the modeling of the random effects is dictated by the design (like a RCB design), and therefore should not change the model effect specifications. If there is only one random effect with 0 estimated variance, one approach is to re-specify that random effect with the statement -- repeated / subject=<random_effect> type=cs;  If the default ddf value becomes inappropriate when a random effect is removed, then you might use ddf= option to specify your own denominator degrees of freedom value, or use other estimation method such as ddfm=kr if that helps. Jill ... View more

G matrix not positive definite might not be devastating, but it is always a good idea to take care of this so you have a clean convergence. If you take out the random effects that have 0 estimated variance, this message is likely going away. The estimates for other effects might remain the same. ... View more

0.9730 in the Covariance Parameter Estimates table is not an element in the G matrix. It is a parameter in the R matrix. It is the correlation in the residuals between two adjacent observations. See the documentation below for more information on the ARH(1) structure. The estimated value for rho is 0.9730. https://go.documentation.sas.com/doc/en/pgmsascdc/v_023/statug/statug_mixed_syntax14.htm#statug.mixed.repeatedstmt_type Also, your G matrix is not positive definite. You might want to take out some of the random effects in your RANDOM statement. Hope this helps, Jill ... View more

I do not think it is possible to estimate separate variance for each subject. If you want to estimate different variance for each gender (valued 0/1), you might do something like the following --   Proc nlmixed  maxiter=10000; Parms b0=1 , b1=0 , b2=0 , b3=0 , a0= 1, g0=1 , V_u0=1, V_u1=1; Aialpha=a0; Gigamma=g0; Bibeta= b0 + b1*x1 + b2 *x2 + b3*time; Lambda=exp(b0 + b1*x1 + b2* x2 + b3*time + +u0+u1*gender);   P1= exp(Aialpha) / (1+ exp(Aialpha) + exp (Gigamma)); P2= exp(Gigamma) / (1+ exp(Gigamma) + exp (Aialpha)); P3=1-p1-p2;   If y=0 then ll = log (p1 + p3*exp(-lambda)); If y=1 then ll= log (p2 + p3*(1)*exp(-lambda)* lambda); If y>1 then ll= log(p3) – lambda + y*log(lambda) – lgamma (y+1);   Model y ~ general (ll); Random u0 u1 ~ normal ([0,0], [v_u0, 0, v_u1]) subject=id; Run;   The random effects estimates are by definition the empirical Bayes estimates. You can use the OUT= option in the RANDOM statement to get these estimates in the OUT= data set.   Hope this helps, Jill ... View more

Sorry I missed specifying Time as a Class variable. Yes, you should specify TIME in the CLASS statement. If you want to test for the TIME effect, you could also add Time in your MODEL statement in PROC MIXED. ... View more