Skip to content

rstandard does not produce standardized residuals

2 messages · Ulrich Halekoh, Brian Ripley

Ulrich Halekoh
# 
Dear all,


the application of the function rstandard() in the base package
 to a glm object does not produce  residuals standardized to 
have variance one:

the reason is that the deviance residuals are  divided
by the dispersion estimate and not by the
square root of the estimate for the dispersion.

Should the function not be changed to produce residuals
with a variance about 1?


R 1.8.1 on windows 2000

ulrich


==============================================================
Ulrich Halekoh,  PhD                         Phone: +45 8999 1825
Biometry Research Unit                       Fax:   +45 8999 1300
Danish Institute of Agricultural Sciences    E-mail: ulrich.halekoh at agrsci.dk
Research Centre Foulum, DK-8830 Tjele, Denmark
Brian Ripley
# 
This has laready been changed in the R-devel version (1.9.0 to be).
Pre-test versions of that for Windows are available on CRAN.
o	rstandard() was wrongly scaled for cases where
	summary(model)$dispersion != 1.

and that ought to mention glm somewhere ....
On Tue, 24 Feb 2004, Ulrich Halekoh wrote: