Statalist


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

Re: st: Fixed effects regressions with probability weights


From   "Leah K. Nelson" <lknelson@ucsd.edu>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Fixed effects regressions with probability weights
Date   Wed, 11 Mar 2009 12:21:04 -0700 (PDT)

Thanks to both Austin and Martin for their suggestions.  As it turns out,
the outcome I am considering is binary.  Any thoughts on conditional
logit-type estimation in which the probability weights vary within groups
(villages)?

Also, in general does using fixed effects estimation automatically cluster
at the level of the fixed effect?

> Leah K. Nelson <lknelson@ucsd.edu>:
>
> You can switch to -areg- which allows pweights that vary within
> village, and produces identical results to using -regress- and
> including a dummy for each village (but one).  How many villages are
> there?  You may want to cluster at the village level:
>
> areg y x [pw=weight_household], a(village) cl(village)
>
> On Mon, Mar 9, 2009 at 4:22 PM, Leah K. Nelson <lknelson@ucsd.edu> wrote:
>> I currently trying to run a regression with fixed effects using
>> probability weights.  The data is in a cross section.  The unit of
>> observation is household, and I am running fixed effects at the village
>> level.  Sampling (probability) weights are assigned at the household
>> level.
>>
>> When I use the command
>>
>> xtreg y x [pw=weight_household], fe i(village)
>>
>> I get an error that says probability weights must be constant within
>> each
>> village.
>>
>> Is this a limitation of Stata, or is there an underlying econometric
>> issue
>> with using household-level probability weights and village-level fixed
>> effects?  If this is simply a Stata issue, is there a way to get around
>> this restriction using the xtreg command?
>>
>> I have already constructed village dummy variables and run the
>> regression
>> including the dummy variables as follows:
>>
>> xi i.village
>>
>> reg y x _I* [pw=weight_household]
>>
>> but I want to make sure that this method (including the fixed effects at
>> the village-level and the probability weights at the household-level)
>> does
>> not yield biased or inconsistent estimates.  Does anyone have any
>> insight
>> on this?
>
> *
> *   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/
>


*
*   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