In this instance, you're right. However, this is a simple model. The
attraction of the latter approach is that it explicitly models the paired
observations. You're telling the model there are 200 observations, but to
compute the standard errors on 100 "clusters". Your original approach
doesn't account for paired design. Controlling for baseline in the first
approach may give you an adjusted "point estimate", but the standard errors
may be biased towards being too narrow. As your investigation becomes more
complex (e.g., with additional covariates), the standard errors may be too
precise in your first approach, and reject the null too frequently.
Paul
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Ricardo Ovaldia
Sent: Friday, October 24, 2003 10:24 AM
To: [email protected]
Subject: RE: st: RE: Stata's logistic vs. SAS CATMOD WLS model.
Thank you very much Paul. It seems to me that this
model is telling me the same as my original model but
in a more convoluted way. In my original logistic
model I included the before measurement as a RHS
variable and the after measurement as the outcome. In
that model -intervention- was significant which I
interpreted as the intervention has an effect on the
outcome. In this new model -intervention- is not
significant, however the -intervention*time- is, but
the conclusion is the same. Is this correct, or am I
missing something?
I really appreciate your time and your help.
Best,
Ricardo.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
