Then I'd say, your villages are your strata, and then your PSUs are
indeed families. For the third village however, you will have to use
finite population correction saying that you sampled some 80% of the
population. If your sampling fractions were high in the first two
villages as well, then you might want to use that, too. I would
imagine your standard errors will actually go down because of those
design effects.
On 8/22/06, Jennifer Leavy <[email protected]> wrote:
Hi
Thanks so much for your input, both Mark and Stas. I wondered if I'd have to go back to the survey design to do this, using some variation on the -svy commands. We designed the survey, three villages were purposively sampled. In two of the villages random samples of households were taken, and where possible one adult male and one adult female surveyed in each household, although this was not always possible for a variety of reasons. For the third village we aimed for full enumeration, though we didn't quite achieve this, and again where possible one male and one female surveyed within each household. So the weights at that level are straightforward, and I did this with earlier analyses.
--
Stas Kolenikov
http://stas.kolenikov.name
*
* 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/
