Clustering at the higher level will take account of lower-level
clustering, for the most part, so you can just:
svyset province [pweight=pw]
svy: reg y x
* or
reg y x [pw=pw], cl(province)
On Sun, May 31, 2009 at 4:06 PM, [ISO-8859-1] Fernando Terrés
<[email protected]> wrote:
> Thank you very much for the answers,
> What I would like to do in a first instance (apart from descriptives and
> totals) is to compare subpopulations with respect to certain variables
> (working conditions), so confidence intervals are needed in order to
> asses if the differences are statistically significant (I didn't knew
> that SPSS had a survey module, but anyway I will use Stata 10 for the
> analysis).
> My problem then is on how to process this data. Talking Stata:
> - Ideally I would have to:
> svyset mun [pweight=pw] || sect || _n, strata(strat)
> With: mun=identifier for municipalities, the included weights,
> sect=identifier for sections, and
> egen strat=group(gender region size activity)
> But: the identifiers are not included in the data set, I'm also unsure
> on whether the stratification is done at the individual level (they say
> that select them by quotes), it seems that strata are defined as
> grouping on gender*region*size*activity because they tabulate those
> values (but in the data file there are some of this groups, a few, with
> different pw, probably errors).
> - If I have correctly understood, Michael suggests doing:
> svyset, poststrata(strat) postweight(pw)
> - On the question of Austin, yes I have an in-between variable, the
> province (50). But I don't understand how to proceed afterwards. One
> cluster with the weights, and then individuals?
> Fernando
>
>
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
