Re: st: Using weights for Pearson correlations (statalist-digest V4 #2044)

[Nick Winter <nw53@cornell.edu>]

Hi all,

There is a Stata FAQ on correlations and survey data, at:

http://www.stata.com/support/faqs/stat/survey.html

The technique recommended there is the one that I implemented in corr_svy, 
which you can download from SSC.

You must use version 8.2 with svyset, because corr_svy does not recognize 
the version 9 svyset.

--Nick Winter


At 06:59 AM 7/25/2005, you wrote:

> Richard Williams <Richard.A.Williams.5@ND.edu> suggested the following code:
> > <snip>
>
>
> > . webuse nhanes2f.dta
> >
> > . version 8.2: svyset [pw=finalwgt], clear
> > pweight is finalwgt
> >
> > . pwcorr female black height [aw=finalwgt], sig
> >
> >               |   female    black   height
> > - -------------+---------------------------
> >        female |   1.0000
> >               |
> >               |
> >         black |   0.0157   1.0000
> >               |   0.1102
> >               |
> >        height |  -0.7041  -0.0041   1.0000
> >               |   0.0000   0.6800
> >               |
> >
> > . corr_svy female black height, sig
> >
> > Survey Correlation
> >
> > pweight:  finalwgt
> > Strata:   <one>
> > PSU:      <observations>
> >
> > Number of observations: 10337
> >
> >               |      female       black      height
> > - -------------+------------------------------------
> >        female |      1.0000
> >               |
> >               |
> >         black |      0.0157      1.0000
> >               |      0.1741
> >               |
> >        height |     -0.7041     -0.0041      1.0000
> >               |      0.0000      0.7185
> > - --------------------------------------------------
> > Key: Estimated Correlation
> >       Significance Level
> >
> > Exact same correlation estimates, but different estimates of the
> > significance levels.  It isn't obvious from these particular printouts, but
> > in general specifying pweights produces the same results as specifying
> > aweights and robust.
> All very interesting and as I would expect, but why is version 8.32 of
> svyset needed?  And why do the correlations change if I use
> version 9?
>
> . webuse nhanes2f.dta
> . svyset [pw=finalwgt], clear
>
>       pweight: finalwgt
>           VCE: linearized
>      Strata 1: <one>
>          SU 1: <observations>
>         FPC 1: <zero>
>
> . corr_svy female black height, sig
>
> Survey Correlation
>
> pweight:  <none>
> Strata:   <one>
> PSU:      <observations>
>
> Number of observations: 10337
>
>              |      female       black      height
> -------------+------------------------------------
>       female |      1.0000
>              |
>              |
>        black |      0.0099      1.0000
>              |      0.3115
>              |
>       height |     -0.6987      0.0057      1.0000
>              |      0.0000      0.5592
> --------------------------------------------------
> Key: Estimated Correlation
>      Significance Level
>
>
>
>
> > =========================
> Paul T Seed
> King's College London,
> Division of Reproductive Health, Endocrinology and Development
>
> St Thomas' Hospital,
> Lambeth Palace Road,
> London SE1 7EH
>
> Funded by the Wellcome Trust & Tommy's the Baby Charity
>
> (Paul.Seed@kcl.ac.uk)
> tel   (+44) (0) 20 7188 3642
> fax (+44) (0) 20 7620 1227
>
>
>
> *
> *   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/
--------------------------------------------------------
Nicholas Winter                           607.255.8819 t
Assistant Professor                       607.255.4530 f
Department of Government              nw53@cornell.edu e
308 White Hall            falcon.arts.cornell.edu/nw53 w
Cornell University
Ithaca, NY 14853-4601

*
*   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/