Re: st: pweight, aweight, and survey data

[Austin Nichols <austinnichols@gmail.com>]

John Westbury <jrwestbury@gmail.com> :
pweights and aweights yield the same point estimates but typically
different variance (SE) estimates; have you read the help files and
documentation available in Stata on weights?  e.g.

[U] 20.18.3 Sampling weights
Sampling weights—probability weights or pweights—refer to
probability-weighted random samples.
Actually, what you specify in [pweight=...] is a variable recording
the number of subjects in
the full population that the sampled observation in your data
represents. That is, an observation that
had probability 1/3 of being included in your sample has pweight 3.
Some researchers have used aweights with this kind of data. If they
do, they are probably making
a mistake. Consider the regression model
...


On Thu, Apr 8, 2010 at 1:42 PM, John Westbury <jrwestbury@gmail.com> wrote:
> Howdy.
>
> I am using a data set generated from a complex sample design.
> Sampling weights are established to account for the probability of
> selection in the sampling design and when applied to records produce a
> nationally representative sample.  Each record in the sample is for
> individuals.
>
> I have experimented obtaining summary statistics with stata weight
> designators of pweight and aweight.  Both pweight and aweight
> designations for my weight variable yield exactly the same summary
> statistics for my data but the definition of pweight and aweight seem
> to differ greatly in stata documentation.  Can someone clarify why
> pweight and aweight yield the same summary statistics?  When would
> using pweight or aweight generate different results?
>
> thanks much
> John

*
*   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/