# Re: st: svyset problem 2: using svy with partially complete surveys

 From sjsamuels@gmail.com To statalist@hsphsun2.harvard.edu Subject Re: st: svyset problem 2: using svy with partially complete surveys Date Fri, 25 Sep 2009 14:46:27 -0400

```Calculating "fpc" for the post-stratified totals is unjustified and
unnecessary: unjustified, because the theory of fpc's applies only to
the original sampling design; unnecessary because the post-stratified
"fpc" values will will be, on average, about same the the design fpc
values.

On Thu, Sep 24, 2009 at 5:46 PM, Michael I. Lichter
<mlichter@buffalo.edu> wrote:
> Peter,
>
> Your question is whether we can or should, in theory, correct variance
> estimates for the size of the sample relative to the population within each
> poststratum, just as is done for "real" strata. It's not a question that
> ever occurred to me before, but it does make sense.  I don't see why this
> could not be done, although it may be mooted by the poststratification
> adjustments to variance calculation that Stata does. Perhaps somebody more
> knowledgeable than me can answer you properly.
>
> Also, you said:
>>
>> An alternative I've considered is to define strata that identify
>> unique combinations of lep and gender and then feeding this
>> information to the poststratification options in svyset.  Problem here
>> is that each PSU, school, now overlaps two strata--one for each gender
>> in that school--and it's not clear what the FPC numbers should be for
>> each strata.  Am guessing this arrangement will probably violate
>> assumptions behind svy.
>
> If you were going to create a poststratum-specific FPC, it would be at the
> student level, not the school level. The original FPC is at the school level
> because that's how you sampled. In creating poststrata defined by student
> characteristics, you are in effect pretending that you sampled students, not
> schools. Your stratum-specific FPC would be at the student level, so
> cross-classification of schools is not a problem.
>
> As for your missing data problem, either using complete cases only or
> following Nick's suggestion about imputation are both probably better
> solutions to your problem than performing some analyses on some cases and
> other analyses on other cases.
>
> Michael
> *
> *   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/
>

--
Steven Samuels
sjsamuels@gmail.com
18 Cantine's Island
Saugerties NY 12477
USA
845-246-0774

*
*   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/
```