Re: st: Re: Weighted regression

[Nick Cox <njcoxstata@gmail.com>]

I guess this was ignored because people found it difficult to identify
a clear question here.

The help for -iweights- (in Stata 10.1, the version accessible to me
as I write) says

' iweights, or importance weights, are weights that indicate the
"importance" of the
 observation in some vague sense.  iweights have no formal statistical
definition; any
 command that supports iweights will define exactly how they are
treated.  Usually, they
 are intended for use by programmers who want to produce a certain
computation.'

You don't say why you want to use -iweights-. It isn't possible to
comment on your report of strange t-statistics, as you don't give any
results and you don't explain why you regard them as strange. It is
not really correct to say that the number of firms (in your case) in
the regression is changed, although what is reported as the number of
observations will change.

That said, the usual weights here would I imagine just be aweights.
But in any case it does no harm and it may be helpful to report
weighted and unweighted regressions.

Nick

On Sun, Apr 3, 2011 at 10:30 AM, mai7777 <mai7777@gmail.com> wrote:
> Any thoughts on this pls?
>
> On Sat, Apr 2, 2011 at 3:19 PM, mai7777 <mai7777@gmail.com> wrote:
>> Hi,
>> I read a bit about the weighting options in stata, awight, iweight...
>> I am running a reg on the stock prices of some firms and need to
>> weight the reg by the market cap of the firms, that is firms with
>> bigger market cap should play a bigger role in the estimation of the
>> betas. iweight changes the number of firms in the regression in order
>> to create the weights, thus it results in strange t-stat results. Is
>> there a way to do an importance weighted regression aside form
>> iweight?
*
*   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/