[R-sig-ME] random effect variance per treatment group in lmer

Alan Cobo-Lewis alanc at umit.maine.edu
Thu Jul 12 00:47:40 CEST 2007

From: "Afshartous, David" <afshart at exchange.sba.miami.edu>
Subject: Re: [R-sig-ME]

"Afshartous, David" <afshart at exchange.sba.miami.edu> wrote:
>(although it seem nonsensical to want such a covariance)


Doug Bates' lmer formulas offer an elegant notation for constraining covariances to zero.

You won't have a covariance parameter if you use
( 0 + as.numeric(drug=="D") | Patient ) + ( 0 + as.numeric(drug=="P") | Patient )

which is why you'd want to use two terms instead of combining them as
( 0 + as.numeric(drug=="D") + as.numeric(drug=="P") | Patient )
because this second version would give you the silly covariace parameter (which unsurprisingly would always have the value 0, even though it would burn a degree of freedom)


Alan B. Cobo-Lewis, Ph.D.		(207) 581-3840 tel
Department of Psychology		(207) 581-6128 fax
University of Maine
Orono, ME 04469-5742     		alanc at maine.edu


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