I have a dataset which I think would suit GEE. I am trying to assess the relative contribution of a number of simultaneous predictors and their interactions to the intensity of hurt feelings. I asked people to report on one hurtful event in each of 3 relationships, as well as fill out a number of scales. I then coded those hurtful events into a number of categories. I have a measure of the person who hurt the participant as a 3 level ordinal RM variable, the type of hurtful event as a 7 level nominal variable, and 2 measures of personality traits (hurt-sensitivity and self-esteem) as continuous variables. I also have a variable which I would like to control for, which is the time since the hurtful event occurred (a continuous variable). (Please let me know if I have explained this dataset clearly.)
I have run a GEE using the 4 main predictor variables (relationship with person who hurt the participant, hurtful event type, hurt-sensitivity and self-esteem) in SPSS, and it looks like it all makes sense and it worked (where ahurtat=intensity of hurt feelings, arellvl=relationship with person who hurt participant, atoe7_rc=type of hurtful event, aHPS_ctrd=hurt-sensitivity and aRSE_ctrd=self-esteem):
GENLIN ahurtat BY arellvl atoe7_rc (ORDER=ASCENDING) WITH aHPS_ctrd aRSE_ctrd /MODEL arellvl atoe7_rc aHPS_ctrd aRSE_ctrd arellvl*atoe7_rc arellvl*aHPS_ctrd arellvl*aRSE_ctrd atoe7_rc*aHPS_ctrd atoe7_rc*aRSE_ctrd aHPS_ctrd*aRSE_ctrd arellvl*atoe7_rc*aHPS_ctrd arellvl*atoe7_rc*aRSE_ctrd arellvl*aHPS_ctrd*aRSE_ctrd INTERCEPT=YES DISTRIBUTION=NORMAL LINK=IDENTITY /CRITERIA SCALE=MLE PCONVERGE=1E-006(ABSOLUTE) SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95 LIKELIHOOD=FULL /EMMEANS TABLES=arellvl SCALE=ORIGINAL /EMMEANS TABLES=atoe7_rc SCALE=ORIGINAL /EMMEANS TABLES=arellvl*atoe7_rc SCALE=ORIGINAL /REPEATED SUBJECT=ID_number SORT=YES CORRTYPE=UNSTRUCTURED ADJUSTCORR=YES COVB=ROBUST MAXITERATIONS=100 PCONVERGE=1e-006(ABSOLUTE) UPDATECORR=1 /MISSING CLASSMISSING=EXCLUDE /PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION COVB.
However, I am not sure how to control for the time since the hurtful event occurred (not included in the above model). I know you can create a Type I model with the GENLIN function, but I can't work out a way to specify that I want the time since the event occurred to be in the first level of the model and the other 4 predictors to be in the second level. Any help would be greatly appreciated.