/ Home

S-Plus Bugs


lm.hat incorrect when weights are present
lm.hat is an undocumented function. Presumably it is intended to do what hat does better, to compute the leverages for a linear model. The computed leverages are not correct however when weights are present.
 
Model extract lm.x ignores subsetting information
model.extract(frame,"lm.x") does not work correctly when there is subsetting information in the frame. It returns the full length model.matrix rather than that with just the subsetted rows. Other requests such as "response" and "weights" work correctly.
 
Named glm families cannot accept variable link

a <- "log"
family <- quasi(link=a)

is ok but

family <- Gamma(link=a)

produces an error. In the latter case, the function Gamma captures the object name "a" instead of the value of the object. The quasi family is the only family which decodes the link correctly.

Arrow Left S Archive.

Arrow Left Statistical Science Web.

Gordon Smyth. Copyright © 1996-2016. Last modified: 10 February 2004