On Mon, Feb 16, 2009 at 4:23 PM, Martin Morgan <mtmor...@fhcrc.org> wrote:
> Stavros Macrakis <macra...@alum.mit.edu> writes:
> > matrix(tapply(mm, outer(rfact,cfact,paste), mean),
> > length(unique(rfact)))
>
> or the variant
>
> idx <- outer(rfact, (cfact - 1) * max(rfact), "+")
> matrix(tapply(m, idx, mean), max(rfact))
>
> The assumption is that cfact, rfact are integer valued with max(rfact)
> <= nrow(m), max(cfact) <= ncol(m).
>
> I think Stavros' solution will run in to trouble when there are more
> than 9 row blocks, and '10 1' sorts before '2 1', for instance.
>
Quite so! Thank you for the correction, and sorry for the sloppy
programming -- using outer/paste like that is indeed a dirty hack.
I suppose the clean way to do this would be to define a cartesian product of
two factors with the induced lexicographic order (is there a standard
function for doing this?):
`*.factor` <- function(f1,f2)
factor( t(outer(f1,f2,paste)),
levels=t(outer(unique(f1),unique(f2),paste)),
ordered=TRUE)
and then I think this will work for arbitrary factors rfact and cfact:
matrix(tapply(mm, rfact*cfact, mean),
length(unique(rfact)))
This allows arbitrary factors, and preserves their input order. Or have I
made another foolish mistake?
-s
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
