Re: st: national survey with complex sample design

[Steven Samuels <ssamuels@albany.edu>]

Some other thoughts. Ordinarily the first-level strata will be  
geographic in a single year, so that "YEAR-GEOGRAPHIC STRATUM" will  
be the highest level stratum. If PSU's are selected within a YEAR,  
there may be independent subsampling in subsequent years.  Thus in  
some surveys, there may be multiple "year" substrata within the first  
level strata.

My point is that Stata can handle these complications as well, if not  
better than, SUDAAN.   Stata, but not SUDAAN, can do sample raking of  
weights to known population totals on multiple control variables,  
thanks to Nick Winter's fine -survwgt- package.

Steve

On Feb 8, 2007, at 2:03 PM, Steven Samuels wrote:

I think that Stata will also work if strata & PSU's do overlap  
years.  For example:  ordinarily "YEAR" would be a first-level  
stratum and PSU's would be sampled independently within a year.   
However if a PSU sampled in one year is re-used for subsequent years,  
then the YEAR of PSU selection would still be the highest level  
stratum, but the year in which the observation is made would be the  
one to use for the analysis of time trends.


Steve


OOn 2/8/07, Hillel Alpert <HALPERT@hsph.harvard.edu> wrote:

> Could someone please advise whether and  how Stata can handle  
> national survey data with multiple years and complex survey sample  
> design?
>
> A single year data can readily be handled in Stata based on the  
> weight and strata variable. Can we analyze the appended data over  
> two or more years the same way, or is a program such as SUDAAN  
> needed? The analysis will be of time trends, and we assume the  
> sampling was not identical from year to year.
>
> Hillel Alpert
> halpert@hsph.harvard.edu
*
*   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/