Re: st: svy: countries as psu

[Steven Samuels <sjhsamuels@earthlink.net>]

I am not knowledgeable about the World Value Surveys. Still, I have  
some comments.

I like Stas's suggestions. It might be difficult to -svyset- the data  
to represent the sampling design below the country levels, but I  
presume, that the WVS "integrated" data sets include unified design  
variables.

I looked at two publications from WVS: one analyzed individuals  
respondent data, with multi-level-models. The other (http:// 
www.worldvaluessurvey.org/Upload/_Liberalism%20and%20%E2%80% 
A6materialism.doc) seems to have used countries as the analysis unit,  
with country-averages for each variable.

You haven't told us much about your purpose and your "target  
population," if any. Your decisions should be based on your purpose.  
If you are interested in analyzing, say, differences between people  
in Easter and Western Europe, you should weight to population. If you  
are interested in differences between countries, possibly you should  
not weight to population.

If you use the supplied weights, then you are effectively weighting  
to population. This will down-weight small countries. if you don't  
use the supplied weights, you risk bias because of some important  
groups within each country are under-represented. See: https:// 
library.usask.ca/files/data/newsletter/datacrunchv13.pdf. Still other  
weighting schemes have been applied to WVS analyses, e.g.: http:// 
www.worldvaluessurvey.org/Upload/_Liberalism%20and%20%E2%80% 
A6materialism.doc

 -Steve

On Oct 23, 2008, at 12:01 PM, Stas Kolenikov wrote:

> As far as I understand, you did not sample the countries. Rather
> whichever participating countries were (a convenience sample, or if
> you are lucky to have this survey in all those countries, a census of
> the country population), the samples were taken independently in those
> countries. So the countries act as (super) strata, and whatever PSUs
> were taken in those countries would be the PSUs to go into the design
> statement. What you are thinking of (the big differences between West
> and East) would be subpopulations, and you can analyze those
> separately using -subpop()- option of -svy- commands.
>
> On 10/23/08, b.todosijevic@utwente.nl <b.todosijevic@utwente.nl>  
> wrote:
> > Dear Stata users:
> >
> >  I need to analyze a comparative data set (World Values Survey),
> >  consisting of a number of country samples, using svy: commands.
> >  I want to obtain various estimated for countries grouped as  
> > Eastern or
> >  Western Europe.
> >
> >  I wonder if it would be appropriate to specify countries as psu's
> >  (clusters), and regions as strata (East-West) in svyset command.
> >  E.g., svyset countries [pw=weight], strata(west).
> >
> >  Thanks,
> >  Bojan Todosijevic
> >
> >
> >
> >  *
> >  *   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/
>
>
> --
> Stas Kolenikov, also found at http://stas.kolenikov.name
> Small print: I use this email account for mailing lists only.
> *
> *   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/

*
*   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/