GEE model effects and parameter estimates don't match

[Putnick, Diane (NIH/NICHD) [E]]

Hello everyone,

I am running a GEE model on a large sample (>300K).  I have a binary DV
(logit link function), 3 continuous covariates, 1 binary within-subjects
factor (parent) and 1 binary between-subjects factor (child gender).  The
model is running fine but I am confused about why the tests of model
effects do not match the parameter estimates for the factors.  They do
match for the covariates, but despite each having 1 df, the Wald's tests
in the parameter estimates section are always smaller than the Wald's
tests in the model effects section for the factors and the interaction
between the factors.  Why is this?


Here is the syntax I'm running:

GENLIN read (REFERENCE=FIRST) BY parent gender (ORDER=ASCENDING) WITH
   cage careu5edu nchilduf
  /MODEL cage careu5edu nchilduf parent gender parent*gender INTERCEPT=YES
    DISTRIBUTION=BINOMIAL LINK=LOGIT
  /CRITERIA METHOD=FISHER(1) SCALE=1 MAXITERATIONS=100 MAXSTEPHALVING=5
    PCONVERGE=1E-006(ABSOLUTE) LCONVERGE=1E-006(ABSOLUTE)
    SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95 LIKELIHOOD=FULL
  /REPEATED SUBJECT=chcareid WITHINSUBJECT=parent 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(EXPONENTIATED)
    HISTORY(1).

I appreciate any insight.

Thanks!
Diane

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD