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, **statalist.org** is already up and running.

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

From |
Steve Samuels <sjsamuels@gmail.com> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: creating variable from svy: proportion |

Date |
Mon, 3 May 2010 12:26:32 -0400 |

You should be fine with these weights. Problems can arise if the original probability weights are post-stratified or otherwise "raked" so that sample estimates better match population totals. These post-stratified weights might not be appropriate for individual cooperatives (or for sub-populations in general). See Levy and Lemeshow, Sampling of Populations, 4th Ed, 2008, page 149. I would have suggested that you use the original probability weights, if available: new_weight = (1/prob sel coop)x (1/prob sel popcode) x (1/ prob sel individual). With these weights, svy: total isb17, over(dccode) would give wrong estimates of totals, because the weight includes a factor for sampling of the cooperative. So, if the probability of selecting a coop was 1/10, the total estimate for that coop would be inflated 10-fold. You only need the product of the last two factors However estimates of proportions and other means will be unbiased, because the extra coop selection factor (10, in the example) cancels out of the numerator (estimate of number of individuals with isb17=1) and denominator (estimate of number of individuals in the coop). Steve On Mon, May 3, 2010 at 11:52 AM, Guy Grossman <guygrossman1@gmail.com> wrote: > weights were constructed by calculating the inverse of the probability > of being included in the sample (basically 1/(multiplying probability > in each of the three sampling stages)). For which constructions you > believe pweights would be biased? what re-weighting method do you > propose? > > Guy > > On Mon, May 3, 2010 at 11:43 AM, Steve Samuels <sjsamuels@gmail.com> wrote: >> How were the weights constructed, Guy? For some constructions, the >> estimates for individual cooperatives could be quite biased, and >> re-weighting might be necessary. >> >> Steve >> >> On Mon, May 3, 2010 at 10:20 AM, Guy Grossman <guygrossman1@gmail.com> wrote: >>> Dear Stata listers, >>> >>> forgive me for posting such a basic question. I am using Stata 10.1 to >>> analyze an original stratified three-stage cluster sample survey in >>> which observation are farmers (individuals) sampled from groups >>> (pocode) themselves sampled from larger farmer cooperatives (dccode). >>> isb17 is a dummy variable measured at the individual level. I am >>> trying to create a variable at the PSUs level (dccode), which is the >>> proportion of farmers in the cooperative that have isb17==1. I am >>> reluctant to use the collapse command given the survey design. >>> >>> svyset dccode [pw= pweight], strata (strata) fpc(fpc1) || pocode, >>> fpc(npos) || _n, fpc(popop) >>> svy: proportion isb17, over(dccode) >>> >>> What I am failing to do, is to take the e(b) from ereturn and generate >>> a variable that represents isb17==1 at the dccode level. >>> Help would be highly appreciated. >>> >>> Cheers, >>> Guy >>> >>> * >>> * 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 >> Voice: 845-246-0774 >> Fax: 206-202-4783 >> >> * >> * 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/ >> > > > > -- > Guy Grossman | PhD Candidate - Columbia University | Political Science > Department | IGERT International Development and Globalization Fellow > | (+1917) 664-6946 | gsg2102@columbia.edu | skype: gugrossman | > http://www.columbia.edu/~gsg2102/ | > > * > * 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 Voice: 845-246-0774 Fax: 206-202-4783 * * 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/

**Follow-Ups**:**Re: st: creating variable from svy: proportion***From:*Guy Grossman <guygrossman1@gmail.com>

**References**:**st: creating variable from svy: proportion***From:*Guy Grossman <guygrossman1@gmail.com>

**Re: st: creating variable from svy: proportion***From:*Steve Samuels <sjsamuels@gmail.com>

**Re: st: creating variable from svy: proportion***From:*Guy Grossman <guygrossman1@gmail.com>

- Prev by Date:
**Subject: st: Stata & LIMDEP comparison for count models** - Next by Date:
**Re: Subject: st: Stata & LIMDEP comparison for count models** - Previous by thread:
**Re: st: creating variable from svy: proportion** - Next by thread:
**Re: st: creating variable from svy: proportion** - Index(es):