R devel - Mar 1999 - Bug in lm.wfit with zero weights in 0.63.3 (PR#136)

R 0.63.3 (any platform)

library(lqs)
example(lqs)

gives

Warning in (y - z$residuals) + offset : longer object length
        is not a multiple of shorter object length

repeatedly. The problem is in lm.wfit, which removes cases with
zero weights, but does not adjust the offset.  The fix is

    z$fitted.values <- (y - z$residuals)
    z$weights <- w
    if (zero.weights) {
	coef[is.na(coef)] <- 0
	f0 <- x0 %*% coef
	if (ny > 1) {
	    save.r[ok, ] <- z$residuals
	    save.r[nok, ] <- y0 - f0
	    save.f[ok, ] <- z$fitted.values + offset[ok,]
	    save.f[nok, ] <- f0 + offset[nok,]
	}
	else {
	    save.r[ok] <- z$residuals
	    save.r[nok] <- y0 - f0
	    save.f[ok] <- z$fitted.values + offset[ok]
	    save.f[nok] <- f0 + offset[nok]
	}
	z$residuals <- save.r
	z$fitted.values <- save.f
	z$weights <- save.w
    } else 
      z$fitted.values <- z$fitted.values + offset
      
The only examples affected are those with a non-zero offset and zero
weights:  these will have incorrect fitted values.

I've commited this for 0.64.0.


-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._