Re: st: how to use subpop with a stratified sample

[Steve Samuels <sjsamuels@gmail.com>]

I know little about dealing with non-probability samples, so take the
comments below with that in mind.

• The subpop() option isn't appropriate for an opt-in survey, as the
theory is based on probability sampling. More important, the effect of the
option is to correct standard errors, but AAPOR states that opt-in
surveys should not present estimates of sampling error.

http://www.aapor.org/opt_in_surveys_and_margin_of_error1.htm

and this prescription is followed by CDC, e.g.

http://www.cdc.gov/mmwr/preview/mmwrhtml/mm6138a1.htm .

So just use an -if- qualifier to analyze subpopulations.

• Stratification weights, post- or otherwise, may not play well with
subpopulation analyses.

For an example, see section 6.4, p 148 of Levy & Lemeshow, 2008.

Levy, Paul S, and Stanley Lemeshow. 2008. Sampling of populations :
methods and applications. Wiley series in survey methodology. Hoboken,
N.J: Wiley.

As you are interested in differences between categories, you might
compare weighted differences with those from a model-based analysis
that ignores the weights and employs the demographic and design
variables, if any, as adjustment factors.

• The singleunit() option in -svyset- is needed only if a stratum has a
single PSU. 

Steve 
sjsamuels@gmail.com


*
*   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/