If you are analyzing a subpopulation, you should use the -subpop-
option in Stata's survey commands. The only exception that I can think
of is, if your analysis is confined to a single, small, subpopulation.
For choosing a centering constant, you have two possibilities: 1) a
simple whole number close to the population mean or median; for
example, if a mean is 102.5, I would use "100" 2) the population mean.
You can estimate this by using -svy: mean-. I recommend the first
choice. If you are analyzing several subpopulations, use the same
centering constant for each, even if the means differ. Otherwise you
will have a difficult time comparing the intercepts. If you are
analyzing only one subpopulation, use a constant close to its mean.
I know nothing about the Add Health data set, but you should check
whether the original sampling weights were post-stratified or "raked"
, so that sample estimates more closely match the population
estimates. If so, and you are analyzing a single subpopulation, you
should consider using only the original sampling weights, not the
post-stratified weights.
Luckily, help is close at hand for making this and other decisions:
the excellent Survey Research Center at Maryland.
-Steve
On Fri, Dec 11, 2009 at 1:28 PM, Castellanos, Karen
<[email protected]> wrote:
> Hi everyone,
>
> I am a PhD student that just found this listserv online. I am using STATA to analyze data from the Add Health dataset for my dissertation. For it, I must enter sampling weights & the primary sampling units to correct for design based effects. Although I know how to center variables, I am uncertain what is the best way to specify my subpopluation when I center. Should I use the subpop option? I am also uncertain regarding whether the sampling weights need to be taken into account and how when centering my variables. Any help you can provide is greatly appreciated.
>
> Thanks,
>
> Karen
>
> Karen Castellanos-Brown, MSW
> PhD Candidate
> Adjunct Instructor
> University of Maryland School of Social Work
>
*
* 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/
