[R] using survey weights for correlations

John Fox jfox at mcmaster.ca
Sun Nov 30 22:00:50 CET 2008

Dear Steve,

You can use svyvar() in the svy package to compute a covariance matrix that properly reflects the weights (and other details of the sampling design), and from this, using cov2cor(), a correlation matrix (if you want that too). You should get consistent estimates from sem() in the sem package (assuming that's what you were planning to use), but standard errors and statistical tests won't be right. You should be able to get valid inferences by bootstrapping, making proper allowance for the weights in resampling.

I hope this helps,

John Fox, Professor
Department of Sociology
McMaster University
Hamilton, Ontario, Canada
web: socserv.mcmaster.ca/jfox

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On
> Behalf Of Steve Powell
> Sent: November-30-08 2:41 PM
> To: R-Help
> Subject: [R] using survey weights for correlations
> Dear list,
> I have a data file which includes, alongside various variables representing
> questionnaire scores, a variable for survey weights computed as the number of
> observations in the sample drawn from that group divided by the number of
> observations in the population in the group. I need to calculate a covariance
> matrix of the questionnaire scores for use in sem. How do I apply the
> weights?
> Thanks in advance,
> Steve Powell
> www.promente.org
> proMENTE social research
> Krančevićeva 35
> 71000 Sarajevo
> skype stevepowell99
> mob. +387 61 215 997
> tel. +387 33 556 865
> fax. +387 33 556 866
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

More information about the R-help mailing list