Greetings,
Thanks to Kit Baum, there are two packages now available from the SSC
archives that deal with replication methods of analyzing complex survey
data.
There are two basic approaches to analyzing complex survey data: Taylor
series linearization (implemented in Stata's svy- commands) and
replication (implemented in these packages). Replication methods
include balanced repeated replication (BRR) and the survey jackknife in
several forms (JK1, JK2, JKn). The replication methods calculate
variance estimates by estimating statistics repeatedly, using multiple
sets of replication weights.
The first package, -svr-, is a suite of programs to conduct analysis
using replication methods. ("svr" stands for SurVey Replication.) It
includes programs, analogous to official Stata's -svy- programs, to
calculate means, totals, ratios, and cross-tabulations. It also
includes -svrmodel-, which estimates regression style models (OLS,
logit, probit, ordered l/probit, etc.); and -svrcorr- which estimates
correlation coefficients. Finally, it includes -svrest-, which
calculates replication-based covariance estimates based on any Stata
command that accepts aweights.
The second package, -survwgt-, is a program that creates replication
weights. It also conducts post-stratification, raking, and non-response
adjustments, so may be of use for analysts using Taylor-series
approaches as well. (Note that -survwgt- is included as part of the
-svr- package, so if you want both you only need to download -svr-).
I've compared output with that generated by WesVar. But, as always, use
at your own risk.
I would appreciate any feedback.
Nick Winter
-----------------------------------------------------------
Nicholas Winter, Ph.D. P 202.939.5343
Policy Studies Associates F 202.939.5732
1718 Connecticut Avenue, NW [email protected]
Washington, DC 20009-1148 www.policystudies.com
-----------------------------------------------------------
<<winmail.dat>>
