predict.lm is broken in 1.1.0-patched (2000-August-7) (PR#626)
ripley@stats.ox.ac.uk writes:
predict.lm has been broken by recent changes to the patched branch. It fails for all singular fits. An example: library(MASS) data(quine) quine.hi <- aov(log(Days + 2.5) ~ .^4, quine) quine.nxt <- update(quine.hi, . ~ . - Eth:Sex:Age:Lrn) predict(quine.nxt) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA
...
I believe this is a lot worse than that which the changes were intended to fix. (Are we not supposed to be ultra-cautious about not making things worse in that branch?)
Sometimes the fix for one bug uncovers another. We have
coef.aov
function (object, ...)
{
z <- object$coef
z[!is.na(z)]
}
coef.lm
function (object, ...) object$coefficients and both of these cannot be right when predict.lm calls coef(object) and object$coefficients contains NA's. Using object$coefficients inside predict.lm fixes things, but I'm still not happy with the two different conventions for coef...
O__ ---- Peter Dalgaard Blegdamsvej 3 c/ /'_ --- Dept. of Biostatistics 2200 Cph. N (*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918 ~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk) FAX: (+45) 35327907 -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-devel-request@stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._