R help - Sep 2002 - subsetting an ordered factor

[I hope all of you had some good holidays/vacation!]

My current flabbergaster:

## a dummy factor
> size<- rep( c('short', 'long'), 10 )
## take a subset:
> subset( size, size!='short')
 [1] "long" "long" "long" "long"
"long" "long" "long" "long"
"long" "long"
## take another subset, all elements that are not some value not actually
found:
> subset( size, size!='small')
 [1] "short" "long"  "short" "long" 
"short" "long"  "short" "long" 
"short" "long"  "short" "long" 
"short" "long"  "short" "long"
[17] "short" "long"  "short" "long" 
## Now order the factor:
> size<-ordered( size, levels=c('short', 'long' ) )
## repeat that subset:
> subset( size, size!='small')
ordered(0)
Levels:  short < long 

So it appears to be invalid/illegal/undefined to ask for all elements in an
*ordered* factor that are *not* some value that does *not* occur in that factor.
Whereas it is perfectly legal to do so for the same, unordered factor?!

## however:
> size<-ordered( size, levels=c('short', 'long',
'small' ) )
> subset( size, size!='small')
 [1] short long  short long  short long  short long  short long  short long 
short long  short long  short long  short long
Levels:  short < long < small 

I have not seen anything in the manpage for subset nor ordered that suggests
this behaviour.

Can someone explain the use of this peculiarity to me, or is this a bug that
should be (has been) reported?

RenE Bertin
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help 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-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

On Thu, 12 Sep 2002 09:08:12 -0400, "Liaw, Andy" <andy_liaw at
merck.com> wrote regarding "RE: [R]
subsetting an ordered factor"

8-) Well, it sort of make sense to me.  An ordered factor has a more defined
8-) domain than an unordered factor.  "!=" for unordered factor simply
compare
8-) the labels, but for ordered factor it should mean "<" or
">".  Thus it
8-) returns NAs because R doesn't know how to do the comparison.


True, I can see it sort of makes sense, but is it a lot of sense? Isn't what
you call an ordered
factor in fact non-existant in the sense that factors are by default and
implicitly sorted
alphabetically? I see the analogy with "== NaN", but I think there may
be a "problem" with Inf? (I
mean: "!=" exists but either "<" or ">" is
undefined -- to forget Inf == Inf! )
Besides: why should factors be treated as numbers; cf. what happens with
characters:
> kk <- c('a', 'b', 'c')
> kk[1]
[1] "a"
> kk[1]< kk[2]
[1] TRUE
> kk[kk!='d']
[1] "a" "b" "c"
## and:
> kk <- c( 'b', 'a', 'c')
## smaller/greater comparison is not based on the index but inherent to the
elements like with
## an ordered factor:
> kk[1]< kk[2]
[1] FALSE
> kk[kk!='d']
[1] "b" "a" "c"

8-) This behavior is at least compatible with Splus.

That's another argument :) Are you sure R crashes everywhere Splus does?
>8- [= mean, lame joke]
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help 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-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._