GEE and weighted dataset

[Melissa Ives]

Good afternoon listers...

 

I am trying to run GEE on repeated measures for a weighted dataset so that sample sizes are similar in characteristics and have an identical size since one group is 475 (OPU) and the other (MAPU) is 6471.  We have already published descriptions of these group as weighted* and are now are trying to look at outcomes across 3, 6 and 12 month follow-ups using the same data and weights.

 

w_opoc is created as a propensity for the MAPU group to be in the OPU group, then population weighted to result in equal size groups.  Thus the weights range from 0-1, with all OPU members having a weight of 1.

 

However, GEE doesn't use weight variables as it rounds to the nearest whole number, so that means any record with a weight <.5 ends up being dropped from the analysis (i.e. MOST of the MAPU group).  Multiplying by 100 (or 50) overpowers the analysis.

 

A colleague ran the GEE in Stata and I was able to almost replicate that by using SCALEWEIGHTS=w_opoc.  However the N's included appear to be larger than the 950 at intake (2792 total weighted records across wave).

 

I have searched SPSS help and the archives

My question is three fold:

    Is scaleweight doing what I want (comparing two 'equivalent' groups) even though it looks like it is including all records?

        If yes, is there a way to show the final group Ns?

        If no, is there a way in SPSS (or Stata--which my colleague knows, but I don't) to do this?

 

Syntax used with SCALEWEIGHT

(opgrps=group indicator (0,2), xobs=wave indicator (0,3,6,12), sfs8p=scale measure, w_opoc=weight variable 0-1).

 

weight off.

* Generalized Estimating Equations.

GENLIN sfs8p BY opgrps xobs (ORDER=ASCENDING)

 /MODEL opgrps xobs opgrps*xobs INTERCEPT=YES SCALEWEIGHT=w_opoc

  DISTRIBUTION=NORMAL LINK=IDENTITY

 /CRITERIA SCALE=MLE PCONVERGE=1E-006(ABSOLUTE) SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95

  LIKELIHOOD=FULL

 /REPEATED SUBJECT=xsite*xpid WITHINSUBJECT=xobs SORT=YES CORRTYPE=AR(1) ADJUSTCORR=YES

  COVB=ROBUST MAXITERATIONS=100 PCONVERGE=1e-006(ABSOLUTE) UPDATECORR=1

 /MISSING CLASSMISSING=EXCLUDE

 /PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION CORB.

*Article describes the weight calculation:

Subramaniam, G. A., Ives, M. L., Stitzer, M. L., & Dennis, M. L. (2010) The added risk of opioid problem use among treatment-seeking youth with marijuana and/or alcohol problem use. Addiction, 105(4), 686-698.

 

Melissa Ives

 

---

PRIVILEGED AND CONFIDENTIAL INFORMATION
This transmittal and any attachments may contain PRIVILEGED AND
CONFIDENTIAL information and is intended only for the use of the
addressee. If you are not the designated recipient, or an employee
or agent authorized to deliver such transmittals to the designated
recipient, you are hereby notified that any dissemination,
copying or publication of this transmittal is strictly prohibited. If
you have received this transmittal in error, please notify us
immediately by replying to the sender and delete this copy from your
system. You may also call us at [phone] for assistance.