R devel - Mar 2002 - Row-binding factor variables looses the ordered attribute.

Hi everyone,
  I've just notice a problem with ordered factor variables. It appears
that row-binding two ordered factors together looses the ordered
attribute. The following example happens both in R1.3.1 and R1.4.1 (on
RedHat 7.2):
> y <- ordered(gl(3,6),labels=1:3)
> y
 [1] 1 1 1 1 1 1 2 2 2 2 2 2 3 3 3 3 3 3
Levels:  1 < 2 < 3
> data <- data.frame(y=ordered(gl(3,6),labels=1:3))
> data[,1]    
 [1] 1 1 1 1 1 1 2 2 2 2 2 2 3 3 3 3 3 3
Levels:  1 < 2 < 3 
> rbind(data,data)[,1]
 [1] 1 1 1 1 1 1 2 2 2 2 2 2 3 3 3 3 3 3 1 1 1 1 1 1 2 2 2 2 2 2 3 3 3 3 3 3
Levels:  1 2 3

I've also noticed a change of behavior, between R1.3.1 and R1.4.1, in the
way some covariate names are constructed. This is really a minor comment
but this change in behavior seems unnecessary and complicates the retrieval
of coefficients using the covariate's name. For example, fitting the
following glm model in R1.3.1 results in
> data <- data.frame(y=rnorm(20),cov=gl(4,5))
> glm(y~I(cov!=2),data=data)
Call:  glm(formula = y ~ I(cov != 2), data = data)

Coefficients:
(Intercept)  I(cov != 2)
     0.8009      -0.7547

Degrees of Freedom: 19 Total (i.e. Null);  18 Residual
Null Deviance:      16.34
Residual Deviance: 14.21        AIC: 55.92

whereas with R1.4.1 we now get
> data <- data.frame(y=rnorm(20),cov=gl(4,5))
> glm(y~I(cov!=2),data=data)
Call:  glm(formula = y ~ I(cov != 2), data = data) 

Coefficients:
    (Intercept)  I(cov != 2)TRUE  
         0.4570          -0.7629  

Degrees of Freedom: 19 Total (i.e. Null);  18 Residual
Null Deviance:      24.39 
Residual Deviance: 22.2         AIC: 64.85 

The covariate name "I(cov != 2)" is now (in R1.4.1) appended by the
value
TRUE. I think that this must come from "model.matrix" which must be
transforming a boolean vector into a factor variable (which I would guess
is not really necessary as this is just a binary vector).
  This change of behavior is not really a problem in itself (if it only
concerns the behavior of "model.matrix") but quite unexpected.
Thanks too all,
Patrick.

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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 Thu, 14 Mar 2002, Patrick Lindsey wrote:

[...]
> The covariate name "I(cov != 2)" is now (in R1.4.1) appended by
the value
> TRUE. I think that this must come from "model.matrix" which must
be
> transforming a boolean vector into a factor variable (which I would guess
> is not really necessary as this is just a binary vector).
>   This change of behavior is not really a problem in itself (if it only
> concerns the behavior of "model.matrix") but quite unexpected.
It is in the NEWS file ....

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