SAS Proc Mixed for Repeated Measures Design with G-G and H-F adjustments

[Lyson]

Hi Colleagues,

I have gone through Sphericity Tests in RMD using SAS Proc GLM, and have used SAS Proc Mixed  to analyse RM experiments as well. My question is how to pupulate the Greenhouse-Geisser (G-G) and the Huynh-Feldt (H-F) adjustments in Proc Mixed procedure SAS results if sphericity fails in the data set. Kindly share the code or explanation please.

[sbxkoenk]

Hello,

Does this paper help?

Comparing the SAS GLM and MIXED Procedures for Repeated Measures
Russ Wolfinger and Ming Chang, SAS Institute Inc., Cary, NC
https://support.sas.com/rnd/app/stat/papers/mixedglm.pdf

( Russ Wolfinger is the original developer of PROC MIXED. )

Koen

[Lyson]

@sbxkoenk Thank you so much. I have gone through Wolfinger & Chang's paper. They say I can get sphericity tests results by running both "TYPE=CS" and "TYPE=HF", where CS --> any covariance structure, and HF is Huynh-Feldt adjustment. So I have tried this code:

proc mixed data=D_univ cl ic covtest;
class CCI4 CHCI3 Time Flask ;
model Leakage=CCI4|CHCI3 ;
repeated / subject=Flask(CCI4*CHCI3) type=csh type=HF;
lsmeans CCI4 / pdiff cl adjust=tukey;
lsmeans CHCI3 / pdiff cl adjust=tukey;
run;

I am not sure if I interpreted Wolfinger & Chang very well by "running both TYPE=CS and TYPE=HF" correctly. Kindly check and assist before I falsely conclude mu results here!

[SteveDenham]

One of the reasons for using PROC MIXED is that the F tests are valid whether or not the sphericity assumption is met, provided a valid covariance structure is fit and that there are no issues with the convergence.  If you are truly concerned about sphericity, then you should stick with PROC GLM, and not use ML/REML methods.

SteveDenham

[Lyson]

@SteveDenham I get your point. I got a way to test for sphericity in Proc MIXED (from the posted source by Wolfinger). My next question is: If sphericity fails, how do we apply the H-F adjustment in Proc MIXED?

[SteveDenham · Accepted solution]

Specifying TYPE=HF in the REPEATED statement can accomplish this.  However, I suspect you missed the point I was trying to make.  Consider other covariance structures - I really doubt that HF will lead to the smallest corrected AIC - it will probably not be the best fit for your model.  This avoids making a sphericity assumption, which is really not necessary (see the chapter on Repeated Measures in any edition of SAS for Mixed Models for a discussion).

SteveDenham