[R] Calculating/understanding variance-covariance matrix of logistic regression (lrm $var)

Martin Maechler maechler at stat.math.ethz.ch
Thu Jan 29 09:22:31 CET 2004

>>>>> "Karl" == Karl Knoblick <karlknoblich at yahoo.de>
>>>>>     on Thu, 29 Jan 2004 02:34:27 +0100 (CET) writes:

    Karl> Hallo!
    Karl> I want to understand / recalculate what is done to get
    Karl> the CI of the logistic regression evaluated with lrm.
    Karl> As far as I came back, my problem is the
    Karl> variance-covariance matrix fit$var of the fit
    Karl> (fit<-lrm(...), fit$var). Here what I found and where
    Karl> I stucked:

    Karl> -----------------
    Karl> library(Design)


The usual ("official") R (and S) way for this is using  
    r <- glm(..., family = binomial)
with  predict(r, .., se.fit=TRUE) 
and   vcov(r)
giving the variance-covariance matrix,
calling the vcov.glm(.) method in this case, which it self
mainly relies on summary.glm(.).


As you see yourself,  lrm() is from a particular CRAN package by
Prof Frank Harrell and if you really want that, you should ask
the package author -- as you are told in the posting guide
(you should read! -- see the last line of every R-help message).

Martin Maechler <maechler at stat.math.ethz.ch>	http://stat.ethz.ch/~maechler/
Seminar fuer Statistik, ETH-Zentrum  LEO C16	Leonhardstr. 27
ETH (Federal Inst. Technology)	8092 Zurich	SWITZERLAND
phone: x-41-1-632-3408		fax: ...-1228			<><

More information about the R-help mailing list