ANCOVA repeated measures

[Giul]

I would like to test if neuronal state predicts behavioral output, and if the slope of the regression changes when taking into account age group (young or elderly) and time of measurement (repeated measure: T1, T2, T3, T4, T5, T6, T7, T8, T9).

Neuronal state and behavioral output are continuous variables, age group and time categorial.

In SAS manual, I found the Analysis of covariance based on GLM procedure, but was unable to indicate the repeated option, as for a PROC MIXED.

Which approach should I use to test this model?

1.)

ODS GRAPHICS ON ;

PROC GLM DATA = TestExcit PLOTS=all;

CLASS group  time; 

MODEL behavior = neuronal   group   time   group*time    neuronal*group*time     /solution;

RUN;

ODS GRAPHICS OFF;

I evaluated the main factors and interactions through Type III SS., and it gave me a display of the 9 slopes for the young and the 9 slopes for the elderly.

2.)

PROC MIXED DATA= TestExcit;

CLASS group  time  subj;

MODEL behavior = neuronal   group   time    group*time    neuronal*group*time     /DDFM = KR;

RANDOM subj;

REPEATED time /TYPE =ar(1) SUBJECT = subj GROUP= group  ;

LSMEANS group   time    group*time    / Adjust=Tukey PDIFF CL;

ods output diffs=diffcm;run;

I evaluated the main factors and interactions through the Type III SS.; however, I am not interested in the LSMEANS output in this case (LSMESTIMATE instead?)…

Feedback are welcome. Many thanks in advance!

[pau13rown]

you may want a random coefficients model. A random coefficients model is useful when the relationship of the measurement with time is of interest, as it seems to be here. You would use the random statement in proc mixed ie "random int time / subject=patient type=UN". This is my sense of it anyway. I'd recommend brown & prescott's book: applied mixed models in medicine, or the book 'sas for mixed models' likely covers it