For repeated measures data, you can use PROC MIXED/PROC GLIMMIX to model the covariance in the data, and the tests for the fixed effects would account for the covariance. This is done in one-step in these procedures. You do not need to have so called covariance corrected original data then use that for subsequent ANOVA analysis. Thanks, Jill ... View more

I am not sure you mean by covariance corrected original data. Can you please clarify? Thanks, Jill ... View more

PROC NLMIXED can definitely fit more than one random effect, just not as how it is done in your model. For example, you could have one random intercept and one random slope -- mu = b0*D0+b1*D1+b2*D2+b3*D3+b4*d4+u1+u2*D1; (just to illustrate the syntax, not saying you should use this model) not 5 random intercepts as in your model -- and that might be the reason why the results do not work for you -- the model is not specified appropriately. I am not aware of a way to fit the AR(1)+CS covariance model in SAS/STAT procedures, unfortunately. The closest I can think of is to use the type=sp(lin) in the REPEATED statement in PROC MIXED with the LOCAL option. Thanks, Jill ... View more

Your PROC NLMIXED model -- mu = b0*D0+b1*D1+b2*D2+b3*D3+b4*d4+u1+u2+u3+u4+u5; has 5 random effects (u1 to u5) for a random intercept -- that is not appropriate. One is enough. It is very uncommon to model the random effect variance structures using CS and/or AR(1). I suspect you wanted to model the residual variance-covariance structures (rather than the random effects variance-covariance structures) with CS and/or AR(1)? If that is the case, then PROC NLMIXED cannot help you. This procedure assumes independent error structure. I am not sure how the "combination  of CS and AR(1)" looks. But one general covariance structure in PROC MIXED is type=lin(q). It specifies the general linear covariance structure with q parameters. This structure consists of a linear combination of known matrices that are input with the LDATA= option. It would accommodate some custom covariance structures as long as it is a linear combination. But with AR(1), it might not work. Thanks, Jill   ... View more

It might be just what the likelihood estimation methods give you in this case. Not sure if there is anything you can do about it, unless there are numeric issues with these methods for your data/model. If you need confirmation, please send in your data and program. Thanks, Jill  ... View more

If LSMEANS or LSMESTIMATE statement works, you can always add the E option in these statements, then use the CONTRAST statement to reproduce one or more of the estimates of your interest. ... View more

Can you send me the Output? Especially the E option output? Thanks, Jill ... View more

Thank you for the information. Does the following statement work for you -- contrast "Site Impact (Location 1): Treatment 1 (vs. Control)" Location 2 0 Treatment_1 0 0 0 0 1 1 -2 0 Location * Treatment_1 0 0 0 0 1 1 -2 0 0 0 0 0 0 0 0 0 / e; Thanks, Jill ... View more

Then please provide the E option output. -- Jill ... View more

I do not see Location and Treatment_1 in your CLASS statement. Is that what you intended? If after making appropriate changes, you are still experiencing issues, send us the Class Level Information table and the E option output. Thanks, Jill   ... View more

SUBJECT=IntervalNum(UnqTONum(MRPSID))  should be changed to SUBJECT=IntervalNum(UnqTONum MRPSID)  Thanks, Jill ... View more

PROC MIXED and PROC GLIMMIX are two independent procedures. In most cases the two procedures perform similarly, but in some cases, the numerical performance can differ. Have you tried the PARMS statement in PROC GLIMMIX? Also you might want to take out METHOD=RSPL option in the PROC GLIMMIX statement. ... View more

This error message -- "The initial estimates did not yield a valid objective function." can often be dealt with by a PARMS statement. You can use the ODS OUTPUT COVPARMS= statement in PROC MIXED to save the parameter estimates into a SAS data set, then in PROC GLIMMIX, use the PDATA= option in the PARMS statement to read in the initial parameter values. As pointed out by @ballardw , it might be a good idea to include the Log, the Output, and the example data set so we can have a better idea of the situation. For example, depending on your data values, the subject= effect might be subject=id_s_p(block id_s) rather than subject=id_s_p. Also, we can understand better your "irregular" time points situation. Thanks, Jill   ... View more

First, your PROC MIXED code does not seem to be appropriate. It looks to me that you wanted to fit a random coefficients model? If so, TIME should not be a CLASS variable, and you should have the SUBJECT= option in the RANDOM statement. Then with this modified RANDOM statement, your REPEATED statement with TYPE=UN structure is likely over-fitting your data and might need to be removed.   proc mixed data=trial; class idno ; model GLOBALCOG = time WMFA WMH ICV brainvol AGE RACE GENDER/s chisq OUTPM=WORK.PM RESIDUAL; random intercept time / subject=idno; run;   Then you might use approaches illustrated in the following usage note to obtain subject-specific intercept and slope estimates.   37109 - Obtaining subject-specific parameter estimates and tests for a random coefficients model or HLM with PROC MIXED or PROC GLIMMIX   Hope this helps, Jill ... View more