summary.data.frame with compound elements problem (PR#1891)

[b.rowlingson@lancaster.ac.uk]

Full_Name: Barry Rowlingson
Version: 1.5.0
OS: x86 linux
Submission from: (NULL) (130.95.16.114)


Elements of a dataframe may be a matrix or another dataframe. In
this case the summary() method can do bad things:

> x <- data.frame(1:10)
> x$z <- data.frame(x=1:10,y=1:10)
> summary(x)
     X1.10          z                  
 Min.   : 1.00   NULL:Min.   : 1.00    
 1st Qu.: 3.25   NULL:1st Qu.: 3.25    
 Median : 5.50   NULL:Median : 5.50    
 Mean   : 5.50   NULL:Mean   : 5.50    
 3rd Qu.: 7.75   NULL:3rd Qu.: 7.75    
 Max.   :10.00   NULL:Max.   :10.00    
                 NULL:Min.   : 1.00    
                 NULL:1st Qu.: 3.25    
                 NULL:Median : 5.50    
                 NULL:Mean   : 5.50    
                 NULL:3rd Qu.: 7.75    
                 NULL:Max.   :10.00    

 - this comes from summary(x) calling summary(x$z) and
not handling the fact that summary(x$z) has an element for
each column.

 I think this may be a can of worms. There is a difference
between:

 x <- data.frame(1:10,z=data.frame(1:10,1:10))
and
 x <- data.frame(1:10) ; x$z<-data.frame(1:10,1:10)

 In the first case data.frame() will take columns of the z= argument
and add them to x so that x ends up with 3 elements. In the second
case x ends up with two elements (although it prints as three columns
with identical names for $z).

 Should data frames (and matrices) be allowed as elements of data frames
(assuming they have the right number of rows)? If _not_, then "$<-.data.frame"
should either behave like data.frame(z=...) above or disallow it, and if _so_
then 
summary.data.frame needs tweaking to cope.

 [Splus has similar behaviour, except only for data.frames in data.frames, but
that's because in Splus summary(a matrix) is summary of the whole matrix and 
so only has one element. In R the summary() of a matrix is an element for each
column. Now where's the Splus jitterbug?]

Baz


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