st: how to use subpop with a stratified sample

[Laura Gibson <lgibson@asc.upenn.edu>]

Hello -

I have a question about the best use of svy, subpop.

We have a dataset which was collected from an opt-in online survey with quotas for cigarette smoking status (never, past, currently). I used the National Survey on Drug Use and Health (NSDUH) to get expected margins for gender, race, and age for each of these smoking statuses. Then using those margins, I created post-stratification weights for each smoking status using the user added command -ipfweight-. I appended them all into a single weight for use with the whole dataset and then use the following code to set the weights:

svyset pid_n [pweight=wgt18], vce(linearized) singleunit(missing)

Now we are running some analyses on various sub-sets of cases which cross smoking statuses (e.g., comparing beliefs of those who smoked in the past (and of this group, only smoked < 100 cigarettes) with those who smoked currently (and of this group, only smoked > 100 cigarettes)).

In this example, could I limit the weights to those created for the two smoking statuses which are being used: past and currently (i.e., drop the never weights) and then use subpop to limit the sample to those with <100 cigs in the past group and >100 cigs in the currently group? 

(There is an added wrinkle that we redistributed respondents based on their overall tobacco use after we collected the data (e.g., never cigarette smokers who are currently using chewing tobacco were moved into the currently group). Would we be better off making the weights based on the original quotas (based only on cigarette smoking status), even though that's not how we will analyze the data (if we made weights based on cigarette smoking status, I think we would have to then use subpop when we analyze the data based on overall tobacco use)?)

Thanks,
-Laura

----------------------------
Laura Gibson, PhD
Research Director
University of Pennsylvania

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/