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

st: sample weights in regression models

From   "Austin Nichols" <[email protected]>
To   [email protected]
Subject   st: sample weights in regression models
Date   Wed, 1 Aug 2007 14:05:12 -0400

Olalekan Uthman --
It seems this attitude toward the use of weights and other survey
characteristics in regressions is quite prevalent--some of my close
friends and colleagues share it, though I do not.  Perhaps underlying
the anti-weight view is the notion that if you correctly specify your
model then using the weights can only inflate the variance of your
estimates. Most of us who run regressions are not quite so confident
that models are correctly specified in all cases, and we also worry
about correlated errors within cluster, especially in complex surveys.
 Even if both weighted and unweighted regressions are unbiased
estimators of B, the unweighted regressions are more efficient while
weighted regressions are more robust to common violations of
assumptions.  I would recommend you always use the survey commands
(e.g. -svy: logit-) on survey data.

You may find this paper a useful summary:
Phillip S. Kott. "A Model-Based Look at Linear Regression with Survey Data."
The American Statistician, Vol. 45, No. 2. (May, 1991), pp. 107-112.

or the earlier
William H. DuMouchel and Greg J. Duncan. "Using Sample Survey Weights
in Multiple Regression Analyses of Stratified Samples." Journal of the
American Statistical Association, Vol. 78, No. 383. (Sep., 1983), pp.
[earlier copy at]

or the even earlier
L. R. Klein and J. N. Morgan. "Results of Alternative Statistical
Treatments of Sample Survey Data." Journal of the American Statistical
Association, Vol. 46, No. 256. (Dec., 1951), pp. 442-460.

On 8/1/07, Olalekan Uthman <[email protected]> wrote:
> "Use of sample weights is inappropriate for estimating
> relationships, such as regression and correlation
> coefficients".
*   For searches and help try:

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