Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: weights
Austin Nichols <firstname.lastname@example.org>
Re: st: weights
Wed, 15 May 2013 07:27:18 -0400
Hsu-Chih <email@example.com> and Nick:
I think one can go farther and correct at least one misapprehension
before starting with a long list of readings. Unweighted regression is
preferred not "because it is less biased and more consistent than the
weighted analysis" but because in a *correctly specified model* it has
lower variance (i.e. we start with the assumption of no bias and then
show that unweighted regression has lower variance, implying lower
mean squared error). Models are rarely if ever correctly specified in
the real world, outside of a simulation, and so the question is one of
trading off a loss of efficiency when weighting against potential
reductions in bias, a very complex tradeoff. Many economists are
comfortable assuming their model is correctly specified and ignoring
weights, but I am not one of them. I would recommend never ignoring
On Tue, May 14, 2013 at 6:27 PM, Nick Cox <firstname.lastname@example.org> wrote:
> Your general questions on weights are best addressed by starting with
> the sections on weights in [U] and then looking at the [SVY] manual:
> these manuals are included as .pdf documents with modern versions of
> I doubt that even sampling experts on this list (I am not one) can
> make authoritative statements about your data from what you tell us.
> Please do read http://www.stata.com/support/faqs/resources/statalist-faq/#others
> before your next post.
> On 14 May 2013 23:15, Cheng, Hsu-Chih <email@example.com> wrote:
>> Dear Statalist veterans,
>> Stata offers four options of weights: frequency weights, analytic weights, sampling weights, and importance weights. Winship and Radbill (1994) suggest that un-weighted regression is preferred because it is less biased and more consistent than the weighted analysis, but their discussion is applicable only to sampling weights. If the values of weights in my data center around 1 (some values are smaller than 1; some are greater), is it possible that these are still sampling weights? Or, if the weights in the data are analytic or importance weights, what are the properties of using these weights in regression analysis? Any suggestion or direction to read more on this issue is highly appreciate.
* For searches and help try: