1. Try the following in 63.3 y <- rnorm(20) y1 <- y[2:20] y2 <- y[1:19] summary(glm(y1~offset(y2))) summary(glm(y1-y2~1)) Both are the same model, but the first gives the wrong s.e. I don't believe this was a problem in 63.2 2. Besides the strsplit buffer, there appear to be other memory problems causing seg faults. With vsize 12M in batch, I accumulated 9 glm objects in memory, then called one of my RM functions and systematically got a seg fault. After much searching, I simply removed the 9 objects before calling my function, the problem went away. This did not occur in 63.2 but I have changed my function so that it now uses strsplit and modifies function environments. However, in this case, strsplit is not applied to any text that would cause buffer overflow. Jim -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- 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 _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
On Wed, 7 Apr 1999 jlindsey@alpha.luc.ac.be wrote:> 1. Try the following in 63.3 > > y <- rnorm(20) > y1 <- y[2:20] > y2 <- y[1:19] > summary(glm(y1~offset(y2))) > summary(glm(y1-y2~1)) > > Both are the same model, but the first gives the wrong s.e. > I don't believe this was a problem in 63.2raw residuals from a glm with offset were wrong: fixed for 0.64. -- 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 _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
On Wed, 7 Apr 1999 jlindsey@alpha.luc.ac.be wrote:> 1. Try the following in 63.3 > > y <- rnorm(20) > y1 <- y[2:20] > y2 <- y[1:19] > summary(glm(y1~offset(y2))) > summary(glm(y1-y2~1)) > > Both are the same model, but the first gives the wrong s.e. > I don't believe this was a problem in 63.2My apologies: I have mis-edited the patch I applied, working too close to the deadline for 0.64. I have in glm.fit residuals[good] <- z - (eta + offset)[good] # z does not have Change + to - . Oh well, 0.64.1 may get this right. Brian -- 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 _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._