The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Sampling and Weighting Webinar June 2015 » pweight vs iweight
pweight vs iweight [message #5518] Tue, 02 June 2015 16:00 Go to next message
sbecker is currently offline  sbecker
Messages: 1
Registered: March 2014
Location: Baltimore, Md. USA
Member
For analyses, pweight in STATA would seem the appropriate option, but I've received a program from DHS (unmet need I believe it was) which used iweights. Is this because the weight in DHS includes the nonresponse factor so it is not a true probability weight or...?

Re: pweight vs iweight [message #7059 is a reply to message #5518] Tue, 18 August 2015 11:32 Go to previous message
Sarah B is currently offline  Sarah B
Messages: 20
Registered: June 2013
Member
When calculating means, proportions, frequencies, or other straightforward tabulations, pweights and iweights will produce identical results. Iweights involve less processing and so Stata can produce results more quickly with iweights than with pweights, which is why I often use them when I just want weighted proportions.

The meaningful difference between iweights and pweights, as I understand it, is the formula used to calculate SEs, in that pweights produce robust SEs. So while iweights and pweights produce the same means, proportions, and frequencies, the SEs, CIs, and calculations of statistical significance will be different when using iweights vs. pweights. So for any calculations involving uncertainty estimation, pweights are indeed the appropriate option.

Cheers,
Sarah
Previous Topic: De-normalization; original probability weights; level of clustering
Next Topic: Webinar Presentation Available Here!
Goto Forum:
  


Current Time: Mon Jul 22 08:50:32 Eastern Daylight Time 2019