[R] R hang-up using lm

[Göran Broström]

On Wed, Apr 28, 2004 at 11:18:38AM +0200, Göran Broström wrote:
> On Wed, Apr 28, 2004 at 11:46:51AM +0300, Renaud Lancelot wrote:
> > Martin Maechler a écrit :
> > >>>>>>"kjetil" == kjetil  <kjetil at acelerate.com>
> > >>>>>>   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)

while 1.9.0 (Linux):


> glm(x ~ y)

Call:  glm(formula = x ~ y) 

Coefficients:
(Intercept)  
        5.5  

Degrees of Freedom: 9 Total (i.e. Null);  9 Residual
Null Deviance:      82.5 
Residual Deviance: 82.5         AIC: 53.48 

> 
> Göran
> 
> 
> > 
> > Best,
> > 
> > Renaud
> > 
> > -- 
> > Dr Renaud Lancelot
> > vétérinaire épidémiologiste
> > Ambassade de France - SCAC
> > BP 834 Antananarivo 101
> > Madagascar
> > 
> > tél. +261 (0)32 04 824 55 (cell)
> >      +261 (0)20 22 494 37 (home)
> > 
> > ______________________________________________
> > R-help at stat.math.ethz.ch 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
> 
> -- 
>  Göran Broström                    tel: +46 90 786 5223
>  Department of Statistics          fax: +46 90 786 6614
>  Umeå University                   http://www.stat.umu.se/egna/gb/
>  SE-90187 Umeå, Sweden             e-mail: gb at stat.umu.se
> 
> ______________________________________________
> R-help at stat.math.ethz.ch 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

-- 
 Göran Broström                    tel: +46 90 786 5223
 Department of Statistics          fax: +46 90 786 6614
 Umeå University                   http://www.stat.umu.se/egna/gb/
 SE-90187 Umeå, Sweden             e-mail: gb at stat.umu.se