求助：
各位大神，我想请问下mixed model 下变量的顺序对结果究竟有什么影响。同样的数据，我采用不同数需的数据跑模型，在同一个structure下的收敛情况不一致。之前有遇见过不同变量顺序下虽然说都是收敛，但是出来的结果lsmean diff 以及p值都不同，在这边很困惑，按照我的理解不应该啊。如下是我的模型：

ods output lsmeans = a2(keep =aweekn TRTGN estimate stderr probt)
            diffs = b2(where=(aweekn=_aweekn and TRTGN ne _TRTGN));
proc mixed data = test noclprint method = reml alpha = .05;
    class  TRTGN aweekn studyid usubjid  ;
    model anaval = base studyid TRTGN aweekn aweekn*TRTGN  /s ddfm = kr alpha = .05;
    repeated aweekn / subject = usubjid type = UN;
    lsmeans aweekn*TRTGN/alpha = .05 pdiff cl e om = test;
quit;

ods output lsmeans = a3(keep =aweekn TRTGN estimate stderr probt)
            diffs = b3(where=(aweekn=_aweekn and TRTGN ne _TRTGN));
proc mixed data = test noclprint method = reml alpha = .05;
     class  TRTGN aweekn studyid usubjid  ;
    model anaval =TRTGN aweekn base studyid   TRTGN* aweekn /s ddfm = kr alpha = .05;
    repeated aweekn / subject = usubjid type = UN;
    lsmeans aweekn*TRTGN/alpha = .05 pdiff cl e om = test ;
quit;