I am working on a LMM in SPSS but I am not really sure whether I am doing the correct analysis. The study is a within-subject design -- so participants came in 2x to the lab in randomized order and basically completed both conditions. Because they've done both conditions I think I need to insert "condition" into the repeated box in SPSS. Am I right to think that I need to include this in the repeated box? I've talked to a statistician here and we've always left that box empty in SPSS as he said it was not that important but I think I should account for the fact that participants are repeating that measurement (condition) 2x. On the other hand, I've also ran the analysis with the repeated box empty and with including the protocol in it too and I get similar results so I guess -- it doesn't matter? Could someone more experienced in this help me with this? Thank you!!!
Here is the code when I account for the repeated condition variable (so condition is coded as 1 - treatment and 0 as placebo). Happiness is a continuous variable that has been measured multiple times during the day. Experimental time goes from 1 to 500min.
MIXED happiness BY experimentalTime condition
/CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(10) SINGULAR(0.000000000001) HCONVERGE(0,
ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
/FIXED=condition experimentalTime experimentalTime*condition | SSTYPE(3)
/METHOD=REML
/REPEATED=condition | SUBJECT(participant_number*experimentalTime) COVTYPE(DIAG)
/PRINT=G R SOLUTION TESTCOV
/RANDOM=INTERCEPT | SUBJECT(participant_number) COVTYPE(AR1)
/EMMEANS=TABLES(experimentalTime*condition) COMPARE (protocol).
*AIC = 329.
In the general case, whether different model terms should be included to separately control for the effects of participants and repeated measurements is a question that can only be answered by having an intimate knowledge of the data, and by considering the trade-offs (like possibly over-constraining your model) that different choices introduce. I can tell you that in my field (psycholinguistics) I rarely see anyone mention REPEATED statements, despite repeated measures designs being the norm.
However, in your specific case, the choice to use the REPEATED statement appears clearly redundant (and might possibly represent an outright mis-specification of the model). This is because you have already included "experiment time" as a fixed factor. As with all linear modelling, when you include a fixed factor in an LMM, you are not only estimating that factor's effects, but also controlling for its effects when estimating other factors.
- Badges
- Science topic
- Similar topics
- Psychology
- Emotional Psychology
- Happiness