On Wed, Apr 28, 2004 at 12:59:54PM +0200, Göran Broström wrote: > On Wed, Apr 28, 2004 at 11:18:38AM +0200, Göran Broström wrote: > > On Wed, Apr 28, 2004 at 11:46:51AM +0300, Renaud Lancelot wrote: > > > Martin Maechler a écrit : > > > >>>>>>"kjetil" == kjetil <[EMAIL PROTECTED]> > > > >>>>>> on Tue, 27 Apr 2004 19:19:59 -0400 writes: > > > > > > > > > > > > kjetil> On 27 Apr 2004 at 16:46, Raubertas, Richard wrote: > > > > >> Within the last few weeks, someone else reported a similar > > > > >> problem when using the results of tapply in a call to rlm(). > > > > >> Note that the result of tapply is a 1D array, and it appears > > > > >> there is a general problem with using such a thing on the > > > > >> RHS in formula-based modeling functions: > > > > >> > > > > >> set.seed(3) > > > > >> yy <- rnorm(20) > > > > >> gg <- rep(1:10, 2) > > > > >> y <- tapply(yy, gg, median) > > > > >> x <- 1:10 > > > > >> z <- lm(y ~ x) # OK > > > > >> z <- lm(x ~ y) # crashes R > > > > >> > > > > >> (R 1.8.1 on Windows XP Pro) > > > > >> > > > > > > > > kjetil> What exactly do you mean by "crashes R" > > > > > > > > kjetil> Doing this in R1.9.0, windows XP pro, there is no indication > > > > of kjetil> problems. > > > > > > > >nor is there with 1.9.0 or R-patched on Linux, > > > >nor with R 1.8.1 on Linux. > > > > > > > >no warning, no error, no problem at all. > > > >Is it really the above (reproducible, thank you!) example > > > >that crashes your R 1.8.1 ? > > > > > > It does it for me: Windows XP Pro, R 1.9.0 (P IV, 2.4 GHz, 256 Mo RAM). > > > It freezes RGui and a few seconds later, a Windows message appears > > > saying that Rgui front-end met a problem and must be closed. > > > > I had to try it too: No crashes on Win2000 pro (1.8.1) or Linux (1.9.0), > > but (in both cases): > > > > > > > lm(y ~ x) > > > > Call: > > lm(formula = y ~ x) > > > > Coefficients: > > (Intercept) x > > -0.8783 0.1293 > > > > > > > lm(x ~ y) > > > > Call: > > lm(formula = x ~ y) > > > > Coefficients: > > (Intercept) > > 5.5 > > > > i.e., only an intercept estimate in the second case! Surely something is > > wrong!? > > Obviousy, y, generated as above, has an attribute that confuses 'lm', > because > > > lm(x ~ as.vector(y)) > > works as expected. To add further to confusion, R-1.8.1 (Windows): > > > glm(x ~ y) > > Error in model.matrix.default(mt, mf, contrasts) : > invalid type for dimname (must be a vector)
so, > model.matrix(x ~ y) (Intercept) 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 1 10 1 attr(,"assign") [1] 0 but > model.matrix(x ~ as.vector(y)) (Intercept) as.vector(y) 1 1 -0.853357506 2 1 -0.711872147 3 1 -0.228785137 4 1 -0.449739758 5 1 0.173914266 6 1 -0.138766243 7 1 -0.433799800 8 1 0.234183701 9 1 0.002728104 10 1 0.733590165 attr(,"assign") [1] 0 1 AND > rr <- model.matrix.default(x ~ y) Segmenteringsfel [EMAIL PROTECTED]:~$ After a restart and repeating > model.matrix.default(x ~ y) several times, I FINALLY got > model.matrix.default(x ~ y) (Intercept) 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 1 10 1 attr(,"assign") [1] 0 > model.matrix.default(x ~ y) (Intercept) y 1 1 -0.853357506 2 1 -0.711872147 3 1 -0.228785137 4 1 -0.449739758 5 1 0.173914266 6 1 -0.138766243 7 1 -0.433799800 8 1 0.234183701 9 1 0.002728104 10 1 0.733590165 attr(,"assign") [1] 0 1 > model.matrix.default(x ~ y) Segmenteringsfel [EMAIL PROTECTED]:~$ (Don't ask me what's going on: but 'Segmenteringsfel' means 'Segmentation fault':) BTW, this was on Debian testing/unstable; R-1.9.0 -- Göran Broström tel: +46 90 786 5223 Department of Statistics fax: +46 90 786 6614 Umeå University http://www.stat.umu.se/egna/gb/ SE-90187 Umeå, Sweden e-mail: [EMAIL PROTECTED] ______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html