Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

Re: st: weights and pooling question

From   Austin Nichols <>
Subject   Re: st: weights and pooling question
Date   Wed, 8 Jun 2011 16:49:00 -0400

Ladislav Kozak <> :
You should construct your own weights, first deciding what population
or dgp you wish to generalize to, then devoting about a year to
figuring out exactly how to create the perfect weights.
But you will probably not be too far off if you divide weights by the
number of times each distinct person appears (1, 2, or 3) and cluster
by person id, or some higher clustering var.

On Wed, Jun 8, 2011 at 4:28 PM, Ladislav Kozak
<> wrote:
> Thank you for reading my message and any advice you may provide.
> Weights and data pooling.
> I am doing linear regression analysis with three waves (H, M, and R)
> of the British Household Panel Survey.  I am pooling the data to
> increase the sample size and the waves are NOT independent (the waves
> have some observations for the same people).  However, I am not sure
> which or even whether to use weights. A problem, perhaps there are
> others, is that wave H only includes England while M and R include all
> of the United Kingdom.  Is there a way to set three different weights
> after "regress" in Stata or should I just combine the three different
> weights into one new variable and use that in the analysis or should I
> just not use any weights?
> There is a BHPS manual which includes an extensive section about
> weights but does not cover this issue.
> I also read on a very old posting on a BHPS user group that weights
> may not be necessary (page 12), however, I am not sure if this applies
> to my case.
> I am a relative novice and so any help is deeply appreciated.
> all my best,
> Ladislav

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index