proc mixed code errors

[wonderwoman]

Portion of the following editor is producing the classic error statement: NOTE: An infinite likelihood is assumed in iteration 0 because of a nonpositive residual       variance estimate Note the data below is just a sample of it - is fully factoral with all 1-4 repeated time samples present Data atracomp;     Input strain $ comp $ time diam; Datalines; 95        Fus    1    1.825772287 95        Fus    2    1.42601572 95        Fus    3    1.420739244 95        Fus    4    1.456064902 95        Lep    1    1.859838703 Fus
Fus	1580    3    1.549731159
Fus	1580    4    1.422556384
Fus	1645    1    1.414225347
Fus	1645    2    1.589653421
Fus	1645    3    1.530931089
Fus	1645    4    1.459880132
Lep	95	1    1.654991843
Lep	95	2    1.466500938
Lep	95	3    1.421560176
Lep	95	4    1.487391453
Lep	1210    1    1.414225347
Lep	1210    2    1.657105307
Lep	1210    3    1.476058942
Lep	1210    4    1.443722042
Lep	1580    1    1.749761889
Lep	1580    2    1.448447445
Lep	1580    3    1.479019946
Lep	1580    4    1.602472049
Lep	1645    1    1.414225347
Lep	1645    2    1.598117643
Lep	1645    3    1.433091065
Lep	1645    4    1.487279395

;

proc mixed data = atracomp;

    class strain comp time;

    model diam= strain|comp|time;

    repeated time / subject=strain*comp type = cs r rcorr;

    lsmeans comp*strain/slice=comp slice=strain pdiff adj=tukey;

run;

/*same stats but running with un error structure to compare AIC*/

proc mixed data = atracomp;

    class strain comp time;

    model diam= strain|comp|time;

    repeated time*comp / subject=strain*comp type = ar(1) r rcorr;

    lsmeans comp*strain/slice=comp slice=strain pdiff adj=tukey;

run;

ANYONE FIND THE ISSUE IN THIS CODE?  It worked beautifully for a similar data set.

[ballardw]

NOTES are not ERRORS. They do tell you something about your data that you may not have been aware of that may compromise the analysis.
I would look to see if you have an extreme value diam.

[Damien_Mather]

Nothing obviously wrong with the code except..when you get it to run so that you are happy..can you assume equal variances across strain, comp and comp*strain? If not, another adjustment choice might be more appropriate.

Maybe this would get more helpful suggestions if you re-posted to the non-SAS-U-specific statistics discussion forum. The question is a bit too specialised for this forum I think. 

Cheers.

Damien

[BeverlyBrown]

I have moved this discussion over to the SAS Statistical Procedures Community. Experts there should be able to help!

[SteveDenham]

The most likely cause of this NOTE is the presence of a duplicate record someplace in the data.  The example data above got munged during cutting and pasting, so I can't tell if that is the problem.  The second most likely cause is an overparameterized model.  Try attaching a file in good shape regarding the data (it would be nice to have the one that works, and the one that throws the NOTE, for comparison), and maybe we can track this down.

Steve Denham