Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: creating variable from svy: proportion

From   Steve Samuels <>
Subject   Re: st: creating variable from svy: proportion
Date   Mon, 3 May 2010 17:28:44 -0400

 svyset popcode [pw= new_pw], strata(dccode) fpc(f1) || _n fpc(f2)

This assumes that sampling was simple random sampling without
replacement(SRSWOR) at each stage. If it was not at either stage, drop
the fpc for that stage. If you had any "certainty" units at any stage,
then the set-up will be more complex. This advice applies also to your
original three-stage

The lesson here for me is that description of a study design as a
"three stage cluster sample" isn't by itself sufficient to write a
-svyset- statement. I should have asked you for more detail.


On Mon, May 3, 2010 at 4:38 PM, Steve Samuels <> wrote:
> For the analyses you have described so far, you have, in effect,
> defined a new  "population" that consists only of the cooperatives
> that were selected at the first stage. For analyzing this
> "population", the original suggestions I made were incorrect. Here are
> my revised suggestions.
> You should declare cooperative as the new "stratum" and  redefine the
> design as two-stage, not three-stage.  You should also not use the
> -over- option. That assumes that the coops in the are random and will
> calculate standard errors and degrees of freedom incorrectly.
> Whether and how to set the finite population correction for the
> popcode selection stage will depend  on how the popcodes and
> individuals were selected. Were the probabilities of selection of
> popcodes the same within a coop?  If so, specify the fpc to be that
> fraction, say f1, , and ignore fpc's at the stage of individual
> selection. If the probabilities were not constant, then do not specify
> the fpc.
> So here's my suggested code for the analysis of the sampled coops only:
> svyset popcode [pw= new_pw], strata(dccode) fpc(f1)
> To calculate proportions and totals, then:
> levelsof dccode, local(levels)
> foreach l of local levels {
>      di "dccode = `i'"
>     svy: prop isb17 if l ==`i', se ci
>     }
> foreach l of local levels {
>      di "dccode = `i'"
>     svy: total isb32a if l ==`i'
>     }
> I'm curious: I assume that you are also doing analyses to describe the
> entire population.  What's your purpose in doing these analyses that
> describe only the cooperatives in the sample?
> Steve
> On Mon, May 3, 2010 at 3:46 PM, Guy Grossman <> wrote:
>> Dear Steve -
>> Thanks a million for the note. One last question which I hope
>> finalizes the thread. I also need to calculate totals at the
>> cooperative level for continuous variable isb32a. following your
>> advice I set the design to reflect only the prob of selecting ind X
>> prob of selecting a group.
>> new_pw = (1/prob sel pocode) x (1/ prob sel individual).
>> svyset dccode [pw= new_pw], strata (strata) fpc(fpc1) || pocode,
>> fpc(npos) || _n fpc(popop)
>> Howeve when I run the svy: total command I get the following error message:
>> svy: total isb32a , over(dccode)
>> (running total on estimation sample)
>> fpc for all observations within a stratum must be the same
>> r(461);
>> any idea how to fix the error?
>> Again, many thanks to all those contributing to the thread!
>> Guy
>> *

Steven Samuels
18 Cantine's Island
Saugerties NY 12477
Voice: 845-246-0774
Fax: 206-202-4783

*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index