Message-ID: <4AFD26C2.1030502@biostat.ku.dk>
Date: 2009-11-13T09:28:34Z
From: Peter Dalgaard
Subject: dfbetas vs dfbeta
In-Reply-To: <26331704.post@talk.nabble.com>
alphaace wrote:
> Hi, I've looked around but can't find a clear answer to the difference for
> these two? Any help?
>
> Thanks!
One is a scaled version of the other, at least for lm objects:
> stats:::dfbetas.lm
function (model, infl = lm.influence(model, do.coef = TRUE),
...)
{
xxi <- chol2inv(model$qr$qr, model$qr$rank)
dfbeta(model, infl)/outer(infl$sigma, sqrt(diag(xxi)))
}
so it is essentially dividing by s.e.(beta), except that it uses
infl$sigma which AFAICS is the leave-one-out residual sd.
--
O__ ---- Peter Dalgaard ?ster Farimagsgade 5, Entr.B
c/ /'_ --- Dept. of Biostatistics PO Box 2099, 1014 Cph. K
(*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk) FAX: (+45) 35327907