[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: Repost: finding median when using svy?

From   "Stas Kolenikov" <[email protected]>
To   [email protected]
Subject   Re: Repost: finding median when using svy?
Date   Thu, 6 Sep 2007 17:20:40 -0500

The issues that Austin's quick fix is not resolving is (i)
stratification, which in terms of the effect on standard errors is
usually a minor issue compared to clustering; (ii) any of the
post-stratification and non-response adjustments need to be done with
every bootstrap subsample, and (iii) the bootstrap estimate of
variance is biased in small samples (and in survey statistics, it is
the # of PSU per stratum, which is often as low as 2, that counts),
and thus some rescaling needs to be done. For point estimation, having
the weights as he suggested is sufficient.

On 9/6/07, Austin Nichols <[email protected]> wrote:
> Stas is right, and you should read the references he cites.
> Note that the problem is not one of estimating the median (that's
> easy), but of estimating the standard error. To get an incorrect SE,
> you can:
> webuse nhanes2
> prog mymed
>  qreg height [aw=finalw]
> end
> bs, cluster(psu): mymed
> but if you don't care about the SE, you can ignore the clustering
> induced by your survey design and just
> qreg height [aw=finalw]

Stas Kolenikov, also found at
Small print: Please do not reply to my Gmail address as I don't check
it regularly.
*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index