Sarah wrote:
> I am analyzing a case-control study for which the sampling probabilities of
> explanatory variables selectively differ between the cases and the controls.
> I do not need to estimate rates or risk estimates for the population, only
> odds ratios.
>
> I have estimated pweights. In order to account for sampling design it appears
> I can use both svy commands and robust cluster commands in my logistic
> regression. I believe some regression diagnosics are more cumbersome using
> svylogit. What is the benefit of using svylogit versus logit with the rubust
> cluster option?
If the sampling design was not stratified, there is no difference. -logit-
with pweights and cluster will produce the same results as -svylogit-. If the
sample was stratified, -svylogit- can account for different variances across
strata whereas -logit- cannot; the variance estimates from -logit- would be
biased.
Allen McDowell
*
* 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/
