[R-sig-ME] sample weights in lmer

Gasper Cankar gasper.cankar at ric.si
Wed Feb 11 12:31:12 CET 2009

Dear R users,

I found in this mailing list the post from Adam that's been unanswered.
Since I came across same problem, I'll join his kind request for any
help on using sample weights with lmer.

Thanks for any response,

Gasper Cankar. PhD 
National Examinations Centre (Ric)
gasper.cankar at ric.si
tel. +386 1 54 84 682
fax. +386 1 54 84 601

Adam's post:

>From aslez at ssc.wisc.edu  Sat Oct 18 00:09:31 2008
From: aslez at ssc.wisc.edu (aslez at ssc.wisc.edu)
Date: Fri, 17 Oct 2008 17:09:31 -0500 (CDT)
Subject: [R-sig-ME] sample weights in lmer
<4739. at webmail.ssc.wisc.edu>

Fellow R users,

I was wondering whether it is possible incorporate sample weights into
lmer command.  My understanding is that the existing weights option
follows the types of weights used with the lm command; that is, they are
the type of weights you would use when doing something like weighted
squares models.  To the best of my knowledge, sample or design weights
not equivalent to the types of model weights we use for WLS.  I have at
least quasi-confirmed this by comparing lmer results to results produced
using HLM.  Without weights, the results are identical. With weights,
differ.  I know for a fact HLM is using sample weights.

This maybe a relatively simple question, but I just made the switch to
so the answer isn't completely obvious to me.  My thought is either that
there is a way of linking svydesign commands with the lmer routine, or
that theres a way of "tricking" lmer so that it produces sample-weighted
results using the WLS-type weights command.

Any help would be much appreciated.


More information about the R-sig-mixed-models mailing list