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

st: Bootstrap estimate of variance

From   "Doogar, Rajib" <>
To   "''" <>
Subject   st: Bootstrap estimate of variance
Date   Fri, 17 Aug 2007 11:30:58 -0500

I would appreciate any help with the following issue/questions.

Given n (iid ~ N(0,s)) observations on x, I want to test x_i=0, (i, 1..n) (one test for each observation). s is unknown. The n observations come from a "perturbed" environment, raising concerns that the in-sample estimate of variance may not be the correct estimate to use in the test.  Fortunately I also have m observations from a "unperturbed" or "normal" environment available.

I could always use the variance of the m observations to estimate s.

1. Can I do better by using bootstrap, e.g.:
// estimate variance based on random sample, store estimates in bsvar.dta
bs var=r(Var), reps(R) [size(Z)] [seed(S)] [nodots] saving(bsvar, replace): summ var1
// obtain mean of R estimated variances
use bsvar.dta, clear
summ var

2. In particular, would the mean of "var" reported by the final "summ" command in the sequence above be a legitimate bootstrap estimate of the variance of x during the "normal" periods?

Many thanks,

Rajib Doogar,
Department of Accountancy,
The University of Illinois at Urbana-Champaign
1206 S. Sixth Street, Champaign, IL 61820
Ph: 217.244.8083, Fax: 217.244.0902

*   For searches and help try:

© Copyright 1996–2023 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index