R help - Oct 2004 - Subset doesn't drop unused factor levels

a <- data.frame(b = rep(1:5, each=2), c=factor(rep("a",10),
levels=c("a","b")))
levels(subset(a, b=1, drop=T)$c)
# [1] "a" "b"

Is this a bug?

Thanks,,

Hadley

hadley wickham wrote:
> a <- data.frame(b = rep(1:5, each=2), c=factor(rep("a",10),
levels=c("a","b")))
> levels(subset(a, b=1, drop=T)$c)
> # [1] "a" "b"
> 
> Is this a bug?
> 
No, drop = TRUE is not doing what you're thinking it's doing. From 
?subset (R-1.9.1):

    The 'drop' argument is passed on to the indexing method for data
      frames.

I think you're hoping it is passed on to the indexing method for 
factors, which it isn't.

--sundar

hadley wickham wrote:
> a <- data.frame(b = rep(1:5, each=2), c=factor(rep("a",10),
levels=c("a","b")))
> levels(subset(a, b=1, drop=T)$c)
> # [1] "a" "b"
> 
> Is this a bug?
> 
Also, I think you meant:

levels(subset(a, b==1, drop=T)$c)

(Note the double-equals for logical equality)

--sundar

hadley wickham <h.wickham at gmail.com> writes:
> a <- data.frame(b = rep(1:5, each=2), c=factor(rep("a",10),
levels=c("a","b")))
> levels(subset(a, b=1, drop=T)$c)
> # [1] "a" "b"
> 
> Is this a bug?
In some older versions of R (at least older than 1.9.0), there was a
documentation bug in that the help page said that it would drop unused
levels. Nowadays, it says 

    drop: passed on to '[' indexing operator.

which is correct and the docs for [.data.frame

    drop: logical.  If 'TRUE' the result is coerced to the lowest
          possible dimension: however, see the Warning below.

(and that Warning seems to have a typo, but leave that for now...). So
no, it is not a bug, that's not what that argument is for.

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907

> From: hadley wickham
> 
> a <- data.frame(b = rep(1:5, each=2), c=factor(rep("a",10), 
> levels=c("a","b")))
> levels(subset(a, b=1, drop=T)$c)
> # [1] "a" "b"
> 
> Is this a bug?
Don't think so:
> args("[.data.frame")
function (x, i, j, drop = if (missing(i)) TRUE else length(cols) == 
    1) 
NULL

So the `drop' argument is passed to the "[" method for data.frame
(as
documented in ?subset), and not the "[" method for factor, as
that's never
called.

Andy
 
> Thanks,,
> 
> Hadley
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! 
> http://www.R-project.org/posting-guide.html
> 
>

hadley wickham wrote:
> a <- data.frame(b = rep(1:5, each=2), c=factor(rep("a",10),
levels=c("a","b")))
> levels(subset(a, b=1, drop=T)$c)
> # [1] "a" "b"
> 
> Is this a bug?
> 
> Thanks,,
> 
> Hadley
> 
This is always controversial.  I am apparently in the small minority in 
believing that the default behavior is what you are wishing for.  That's 
why the Hmisc package by default drops unused levels (but allows you to 
override that with options(drop.unused.levels=FALSE).  It is distasteful 
to have to override system behavior but I felt I had to in this case. 
No one in R-core wanted to add a non-default option to R e.g. 
options(drop.unused.levels=TRUE).

Frank

-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University