[R] arrays emerging from tapply

[Peter Dalgaard]

"BXC (Bendix Carstensen)" <bxc at steno.dk> writes:

> The code below illustrates some points about results from tapply that
> I find strange. I wonder if they are intended and if so why it is so.
> 
> 1) When you make a table the dimnames is a *named* list, tapply
>    returns an unnamed list.

That depends on how you call it:

> tapply(airquality$Ozone,list(m=airquality$Month), mean,na.rm=T)
m
       5        6        7        8        9
23.61538 29.44444 59.11538 59.96154 31.44828

The table() function does a bit of extra user-friendliness using the
code controlled by deparse.level, that's all. If we had a named.list()
(or so) function that extended list() with similar functionality,
tapply calls could be made almost as user friendly.

 
> 2) data.frame behaves differently on an array and a table. Is this
>    an intended feature?

Yes (of course it is. What did you expect? Someone to have coded this
in his sleep?). You do want to be able to convert a matrix to a data
frame in the obvious way (d[i,j] <- m[i,j]), but for a table it is
more likely that you want to convert to the format suitable for a
Poisson glm analysis.
 
> 3) For tables class(TAB) and attr(TAB,"class") both return "table",
>    but class(ARR) returns "array" whereas attr(TAB,"class") returns
>    NULL.

Yes, that's a design quirk. In S3, objects only had a class if
explicitly specified via a class attribute, but nowadays all objects
have a class, so that methods can be dispatched on "numeric", etc.

The remaining issue (even if you don't mention it) is why tapply does
not return a table. This could arguably be a better choice, but I
suspect that some code depends on using tapply in ways that would be
incompatible. The change to returning a 1D array instead of a vector
certainly caused some trouble. 

-- 
   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