Re: st: RES: multistage sample survey and coding

[Steven Samuels <sjsamuels@gmail.com>]

Ruwan

fpc's should usually be employed only for descriptive statistics about the population that you sampled. If the purpose of your study is analystic: to  assess associations, fit models, compare p or otherwise test hypotheses, you should *omit* all finite population corrections.  

The reasoning is interesting (Cochran, 1977, p.39): It is seldom of scientific interest to ask if a null hypothesis (e.g. that two proportions are equal) is exactly true in a finite population . Except by a very rare chance, a null hypothesis will never be true. You would discover this by enumerating the entire population. This leads to the adoption of a "superpopulation" viewpoint, which is taken by almost all statisticians these days. See also Deming(1966) pp 247-261 "Distinction between enumerative and analystic studies"; Korn and Graubard (1999), p. 227. 

The omission of the fpc turns the sampling specification from sampling of PSUA without replacement (WOR) into sampling with replacement (WR). In WR sampling there is no need to specify sampling stages beyond the first, because the standard error automatically includes contributions from those stages.

Some other comments:
1. You have not weighted your data.
2. Your -singleunit- specification will throw away information.  You should study the -singleunit0 options and see which one best fits your design.
3. I second Billy's suggestion to consult a good text. Billy referred to Heeringa, West, and Berglund (2010) The best about analysis issues (like the single units), in my opinion, is the one by Korn and Graubard (1999).

Steve

References
Cochran, W. G. (1977). Sampling techniques (3ded.). New York: Wiley.

Deming, W. E. (1966). Some theory of sampling. New York: Dover Publications. 

Heeringa, S., West, B. T., & Berglund, P. A. (2010). Applied survey data analysis. Boca Raton, FL: Chapman & Hall/CRC

Korn, E. L., & Graubard, B. I. (1999). Analysis of health surveys (Wiley series in probability and statistics). New York: Wiley.
  


On Sep 24, 2011, at 12:21 PM, William Buchanan wrote:

Hi Ruwan,

You should take a look at the section of the manual that discusses Finite Population Correction.  There is also a website for the book "Applied Survey Data Analysis" from the folks at U Michigan that has tons of great information and useful Stata examples that can help you out.  The authors of the book are also excellent with getting back to people quickly about questions like this.  But if you can determine the FPC at each of the stages it should prevent Stata from assuming that the subjects were sampled with replacement.

Billy Buchanan

Sent from my iPhone

On Sep 24, 2011, at 8:01, "Henrique Neder" <hdneder@ufu.br> wrote:

> Ruwan
> 
> Perhaps the example 6 on page 147 [SVY] Survey Data Manual Stata version 11
> is a good answer for your question.
> 
> I performed an unreal example with 3 stages:
> 
> sysuse auto
> 
> svyset length, fpc(mpg) vce(linearized) singleunit(missing) || length || _
> 
> Note: stage 2 is sampled with replacement; all further stages will be
> ignored
> 
>     pweight: <none>
>         VCE: linearized
> Single unit: missing
>    Strata 1: <one>
>        SU 1: length
>       FPC 1: mpg
>    Strata 2: <one>
>        SU 2: length
>       FPC 2: <zero> 
> 
> 
> Henrique
> 
> -----Mensagem original-----
> De: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] Em nome de r
> Enviada em: sábado, 24 de setembro de 2011 05:43do 
> Para: statalist@hsphsun2.harvard.edu
> Assunto: st: multistage sample survey and coding
> 
> Dear Statalisters,
> 
> I am new to multistage sampling and surveys with Stata and would like
> to get a sense if I am on the right track with my coding.
> 
> I have survey data from a three stage cluster survey that demarcates
> the following stages: District (1), Village (2), Household (3). The
> districts and villages were both selected proportional to population
> sizes.
> 
> Would this be the correct coding for setting the parameters of the
> survey design?
> If so, why is the second sampling unit (village) not included in svy
> description of the sampling stages?
> 
> . svyset district, vce(linearized) singleunit(missing) || village|| _n
> 
> (When entered, this following is returned:)
> 
> Note: stage 1 is sampled with replacement; all further stages will be
> ignored
> 
>     pweight: <none>
>         VCE: linearized
> Single unit: missing
>    Strata 1: <one>
>        SU 1: district
>       FPC 1: <zero>
> 
> Many thanks in advance for your thoughts.
> 
> 
> Ruwan
> *
> *   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/

*
*   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/