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

Re: st: bootstrap and XTIVREG2

From   "Stas Kolenikov" <>
Subject   Re: st: bootstrap and XTIVREG2
Date   Wed, 29 Aug 2007 16:59:31 -0500

On 8/29/07, Austin Nichols <> wrote:
> Erasmo--
> At a minimum, you probably need the cluster() and idcluster() options
> on -bootstrap- (see -help bootstrap-) so you are not resampling from N
> obs, but M clusters (each panel) and there is a new id variable to
> supply to your cluster() option on -xtivreg2-.  Something like:
> bootstrap, r(1000) cl(gvkey) id(g): xtivreg2 `v', first cl(g) fe i(g)
> might work.
> As no doubt Stas Kolenikov will point out, -bootstrap- is not
> guaranteed to solve your problems.

I'll just sob in despair...

Well for this one it may not be that terribly bad, as this is still
some sort of M-estimation/estimating equations setup, and the
regularity conditions such as smooth derivatives may be reasonably
easy to establish, but then I'd be surprised if the bootstrap would
give a result substantially different from -xtivreg2-'s original
sandwich estimator. If it did, then it means the asymptotics is not
there yet for the sample size you have at hand, and then neither is a
satisfactory answer. E.g. if you had 15 clusters then your estimation
procedure is pretty much guaranteed to go astray either way no matter
how many observations you have in each cluster.

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 1996–2020 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index