The difference you observed is caused by the Kackar-Harville correction applied by SAS Proc Mixed when the DDFM=KENWARDROGER option is set. JMP doesn’t apply this adjustment to the BLUP standard errors. For details, please see http://www.jmp.com/support/help/13/The_Kackar-Harville_Correction.shtml#1502602
To get the same results, replace DDFM=KENWARDROGER with DDFM=SATTERTHWAITE in your SAS code.
Dear All,
I am trying to compare the results of SAS PROC Mixed and JMP mixed model. I thought the results would be the same, but they are not. I hope JMP can give the same results of SAS PROC MIXED. Could anyone help me with it?
My SAS code as following:
proc mixed data=Data;
class sample;
model result= time / DDFM=KENWARDROGER s cl E3 alphap=0.05 residual outp=mixedout;
random sample /s;
ods output SolutionF=F SolutionR=Ran ;
run;
The results for the fixed effect are
Effect | Estimate | Standard Error | DF | t Value | Pr > |t| | Alpha | Lower | Upper |
Intercept | 7.2638 | 0.197 | 15.4 | 36.86 | <.0001 | 0.05 | 6.8449 | 7.6828 |
time | -0.4636 | 0.0615 | 14 | -7.54 | <.0001 | 0.05 | -0.5955 | -0.3317 |
The results for the random effect are:
Effect | lot | Estimate | Std Err Pred | DF | t Value | Pr > |t| |
sample | 1 | -0.01193 | 0.1762 | 4.95 | -0.07 | 0.9487 |
sample | 2 | 0.1647 | 0.1762 | 4.95 | 0.93 | 0.3932 |
sample | 3 | -0.2869 | 0.1762 | 4.95 | -1.63 | 0.165 |
sample | 4 | 0.1799 | 0.1762 | 4.95 | 1.02 | 0.3546 |
sample | 5 | -0.04579 | 0.1762 | 4.95 | -0.26 | 0.8055 |
Then, I used JMP's interface to generate the following code:
DATA Data1; INPUT time sample result; Lines;
1 1 7.04755019105043
2 1 5.84207852539687
3 1 6.02392278004353
4 1 5.43660061149986
1 2 7.27729854732895
2 2 6.5688837201289
3 2 5.6460380590458
4 2 5.88401060914374
1 3 6.15741854776455
2 3 5.90111584525169
3 3 5.78059226421567
4 3 4.91383166620419
1 4 6.98283726578696
2 4 6.3260371191925
3 4 6.09507016140813
4 4 6.06037720591234
1 5 6.95244413898138
2 5 6.36002369549625
3 5 5.94090548712508
4 5 4.9001077313589
;
RUN;
PROC MIXED ASYCOV DATA=Data1 ALPHA=0.05;
CLASS sample;
MODEL result = time/ SOLUTION DDFM=KENWARDROGER;
RANDOM sample / SOLUTION ;
RUN;
I got:
Parameter Estimates
Term | Estimate | Std Error | DFDen | t Ratio | Prob>|t| | |
Intercept | 7.2638308 | 0.197046 | 15.45 | 36.86 | <.0001* | |
time | -0.463589 | 0.061501 | 14 | -7.54 | <.0001* |
Random Effect Predictions
Term | BLUP | Std Error | DFDen | t Ratio | Prob>|t| | |
sample[1] | -0.011928 | 0.153244 | 4.948 | -0.08 | 0.9410 | |
sample[2] | 0.1647348 | 0.153244 | 4.948 | 1.07 | 0.3320 | |
sample[3] | -0.28692 | 0.153244 | 4.948 | -1.87 | 0.1207 | |
sample[4] | 0.1799016 | 0.153244 | 4.948 | 1.17 | 0.2938 | |
sample[5] | -0.045789 | 0.153244 | 4.948 | -0.30 | 0.7772 |
It seems like Std error estimates for random effects are not the same in JMP and SAS. I am wondering if there is anyone know how could I code the JMP to make it is the same results as SAS.
Thank you very much,
Anastasia
I think that this kind of question is best served by support@jmp.com. They are very friendly, almost as friendly as the people here!
The difference you observed is caused by the Kackar-Harville correction applied by SAS Proc Mixed when the DDFM=KENWARDROGER option is set. JMP doesn’t apply this adjustment to the BLUP standard errors. For details, please see http://www.jmp.com/support/help/13/The_Kackar-Harville_Correction.shtml#1502602
To get the same results, replace DDFM=KENWARDROGER with DDFM=SATTERTHWAITE in your SAS code.
Thanks for clarification.
I work in environments where both SAS and JMP are in use; therefore I appreciate consistency between both. In case where certain differences are intended by SAS Institute, these should always be indicated in the documentation.