[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
RE: st: how to weight
At 11:07 PM 3/10/2004 +0000, Nick Cox wrote:
Agreed. If the authors have provided weights, they've also hopefully
provided information on how to use them. It might also be that, instead of
just using the weights, you want to use Stata's svy commands (e.g.
clustering and/or stratification may have been used when gathering the
data, and if so variables that indicate that are also hopefully included.)
As you say, non-integers cannot be frequency weights.
Otherwise how these weights are to be interpreted looks like
a substantive matter. You need to move upstream
towards the author or supplier of the dataset.
I would also highly recommend getting the Stata 8 User's Guide and going
over sections 23.14 (Robust variance estimates) and 23.16 (weighted
estimation). These have the best explanations I have seen of Stata's
weighting procedures (does anyone know of anything on the web that comes
close to providing an adequate explanation?). Depending on the data, you
may also want the survey data manual.
The main (only?) difference between pweights and aweights that I am aware
of is that pweights forces the use of robust standard errors, i.e.
parameter estimates are not changed but standard errors are. SPSS, as far
as I know, does not provide an easy means of getting the robust standard
errors, which I would say is one of SPSS's weaknesses.
If analyses have already been done in SPSS using the weight command, my
wild guess is that you will get the same results by using aweights in
Stata. But, my wild guess is also that, either way, the results will be
wrong; the Stata Users guide has a good explanation of how aweights have
been, and continue to be, used improperly. If I had to bet, I would bet on
pweights being best. But again, you really need to know more about the
data before you can decide on how to proceed.
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
WWW (personal): http://www.nd.edu/~rwilliam
WWW (department): http://www.nd.edu/~soc
* For searches and help try: