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

Re: st: Poststratification weights

From   "Michael I. Lichter" <>
Subject   Re: st: Poststratification weights
Date   Wed, 26 Nov 2008 14:10:50 -0500

Thanks, Austin. I should have just run that test myself. I'll try a couple other tests and get back to the list if I see anything unexpected. FWIW :).


Austin Nichols wrote:
Michael I. Lichter <>:
Assuming you specify the same strata for both cases, there should be
no difference, I believe:

*note survwgt on SSC: ssc inst survwgt
webuse poststrata, clear
survwgt post weight, by(type) t(postwgt) replace
svyset, strata(type) poststrata(type) postweight(postwgt)
svy: mean totexp
qui svyset [pw=weight], strata(type)
svy: mean totexp

On Wed, Nov 26, 2008 at 12:38 PM, Michael I. Lichter
<> wrote:
In the Stata svy commands, what is the difference between creating your own
poststratification weight (e.g. using -survwgt-) and treating it as a
pweight (through -svyset [pw=x], strata(y)-) vs. telling Stata (through
-svyset, poststrata(y) postweight(z)-) that you have a postratification
variable and cell count? This shouldn't affect point estimates, but how does
Stata treat the two differently with respect to variance calculations or
anything else? Thanks. (I would RTFM if I had access to one, and I didn't
find a FAQ or any discussion on the topic. I am using Stata 9.2, FWIW.)
*   For searches and help try:

Michael I. Lichter, Ph.D.
Research Assistant Professor & NRSA Fellow
UB Department of Family Medicine / Primary Care Research Institute
UB Clinical Center, 462 Grider Street, Buffalo, NY 14215
Office: CC 125 / Phone: 716-898-4751 / E-Mail:

*   For searches and help try:

© Copyright 1996–2021 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index