Statalist


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

st: probability weighting in -xtlogit, re-


From   Thomas Klausch <thomas.klausch@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: probability weighting in -xtlogit, re-
Date   Wed, 21 Oct 2009 16:21:35 +0200

Dear list members,

The command -xtlogit, re- does not allow for probability weights
(-pweight-), but only importance weights (-iweight-). The mixed
effects version -xtmelogit- does not even allow for any weighting at
all. It has been noted earlier on this list that this might be due to
the fact that methods research is not unanimous about how to handle
weights in categorical multilevel models.

I, however, need to use some sort of weight strategy, since I
individuals were surveyed by a stratified sample. Selection
probabilities of individuals in the strata are known allowing for the
calculation of pweights. This data was matched back to longitudinal
data from a panel. Weights thus are constant across time and stratum.
A binomial dependent variable is to be modelled using a logistic link
function and taking into account the selection probabilities (or
-pweight-).

This however is not possible with -xtlogit, re-. There is the option
to use -iweight-, but I do not know how a suitable
definition/transformation of pweights to iweights would look like. I
wouldn't know any different useful procedure in Stata, so I'd
appreciate any advice on this problem.

Thanks in advance
Thomas
*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index