Multivariate multilevel model simulation produces residuals that are interpreted as estimates of teacher effectiveness.    I ran the multivariate multilevel model as follows: proc mixed data=newdat covtest;    class idt idk cwave;    model y = cwave  / noint s residual outp=Resid;     random cwave / sub=idt type=un g gcorr solution;    repeated cwave / sub=idk(idt) type=un r rcorr;      ods output covparms=cp fitstatistics=fit; run;   I applied independent univariate models as follows:   /*Math*/ proc mixed data=testdat covtest;    class idt idk;    model y1 =  / s residual outp=mathresid;     random intercept / sub=idt type=un solution;    random intercept / sub=idk(idt) type=un solution;      ods output covparms=cpy1 fitstatistics=fity1; run;   proc transpose data=cpy1 out=cpy1_trans; var estimate; id covparm subject; run;   /*Read*/ proc mixed data=testdat covtest ;    class idt idk;    model y2 =  / s residual outp=readresid;    random intercept / sub=idt type=vc solution;    random intercept / sub=idk(idt) type=vc solution;      ods output covparms=cpy2 fitstatistics=fity2; run;   /*Motivation*/ proc mixed data=testdat covtest;    class idt idk;    model y3 =  / s residual outp=motresid;    random intercept / sub=idt type=vc solution;    random intercept / sub=idk(idt) type=vc solution;      ods output covparms=cpy3 fitstatistics=fity3; run;   The deviance from the multivariate model was calculated.   *MM -2LL (deviance); proc phreg data=newdat; class idt idk cwave; model y= ; run;     And then, the deviance of the univariate model was calculated separately.   *UM -2LL (deviance): simplest form no other datat provided; proc phreg data=testdat;    class idt idk;    model y1 = ;   run; proc phreg data=testdat;    class idt idk;    model y2 = idt idk ;   run; proc phreg data=testdat;    class idt idk;    model y3 = ;   run;     The deviance statistics for the univariate model came out to be  3633.48.  The deviance statistics for the multivariate model came out to be 4609.104.     /*LRT = abs(L1–L2);*/   Therefore the absolute difference=975.624.    I need to do the likelihood ratio test (LRT) to compare the difference between the deviances of the models to a chi-square distribution with a degrees of freedom equal to the difference in the number of parameters between models. I am not sure if the following code enables me to do this test.    /*χ2 ((deviance of multivariate)-(deviance of univariate))= deviance of multivariate)-deviance of univariate;*/                 ... View more

Your macro variable TCOV contains curly braces. Those braces are not valid syntax in the SAS DATA step when you make the SUM call. In other words, the macro is resolving to SUM({ .25 .175 .0225, ...}), which is not valid in the DATA step.  even if you remove the curly braces, the positions of the commas in TCOV are not valid for the DATA step SUM function. ... View more

When I look at the data, each teacher has the same 10 students. I want to have the students have unique IDs.How can I have student ID1-student ID150, and have 10 students per each teacher (teacher ID 1-15) at the same time?  ... View more

Thank you for your response. It is very useful. I ran the code and got the following data.  And I got the expected number of observations (150 observations). Each teacher should have the different 10 students, and they each should have their own set of students. Y1, Y2, and Y3 for each ID seem to be different. Does it mean that each teacher has the different 10 students, and they each have their own set of students? But each teacher ID has the same student ID. Does it mean that each teacher has the same10 students, and they each do not have their own set of students? If so, can you please let me know what I can do?  ... View more

When you paste your screen captures into Microsoft Word, they become extremely difficult to read. Please don't do that any more. Just click on the Photos icon and attach your screen capture that way.   It does not appear that you are showing me MMResids or NEWDAT. But, I still want YOU to look at NEWDAT and see what the values of the variable named RESID are in there. You will probably find out that a variable named RESID is not in there. Is that correct? If it is correct, then you can't use NEWDAT to work with variable RESID.   I created Resid by adding "outp=Resid" (Please see the first code).   Okay, now we seem to have found the problem. OUTP=RESID creates a data set named RESID, it does not create a variable named RESID in NEWDAT. You need to merge the residuals in data set RESID into NEWDAT, or maybe just use data set RESID.   My point here is that you can do a lot of this debugging yourself by looking at the SASLOG and looking at the contents of your data sets. You would have seen immediately that variable RESID is not in MMresids and variable RESID is not in NEWDAT. ... View more