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

Re: st: "svyset" for multistage sampling

From   Steven Samuels <>
Subject   Re: st: "svyset" for multistage sampling
Date   Sun, 8 Mar 2009 11:28:47 -0400


My short answer is that your second approach is OK. If you are doing descriptive analyses, you may add the -fpc()- option. However if your goal is to assess association or to test hypotheses, you should omit it. Including the later stages of the sampling design will ordinarily make little difference in the results. The degrees of freedom for estimates will approximately be: d.f. = No. Strata - No. PSU's in either case.

On Mar 7, 2009, at 2:44 AM, Hisako Kobayashi wrote:

Dear Statalist,

I have survey data collected through multistage cluster sampling with
stratification.  I have a question about command of ?svyset.?

For my data, first, the population was stratified and several clusters
were randomly selected for each stratum.  After that, different
sampling methods were used for selected clusters.  For some clusters,
sub-clusters were selected and then for each sub-cluster, individuals
were randomly selected. But for some cluster, individuals were
selected without selecting sub-clusters.  Moreover, some sub-clusters
were stratified and individuals were selected. In short, strata have
different stages and some have within-stratification (like below).

-Stratification ? clusters ? sub-clusters ? sampling
-Stratification ? clusters ? sampling
-Stratification ? clusters ? sub-clusters ? stratification ? sampling

In this case, I think that the command for multistage design like
below cannot be used for my data.
        svyset su1 [pw=pwt], fpc(fpc1)                        ///
               || su2, strata(strata2) fpc(fpc2) || _n, fpc(fpc3)

Instead, I guess I should have the command like...

       svyset [pweight=pwt], psu(su1) strata(strata1)...(without fpc)

I would like to know whether this is correct.
Thank you very much for your advice in advance.


*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index