[R] Compute rank within factor groups

[Peter Dalgaard]

Ken Williams wrote:
> Hi,
>
> I have a data.frame which is ordered by score, and has a factor column:
>
>   Browse[1]> wc[c("report","score")]
>           report score
>   9         ADEA  0.96
>   8         ADEA  0.90
>   11 Asylum_FED9  0.86
>   3         ADEA  0.75
>   14 Asylum_FED9  0.60
>   5         ADEA  0.56
>   13 Asylum_FED9  0.51
>   16 Asylum_FED9  0.51
>   2         ADEA  0.42
>   7         ADEA  0.31
>   17 Asylum_FED9  0.27
>   1         ADEA  0.17
>   4         ADEA  0.17
>   6         ADEA  0.12
>   10        ADEA  0.11
>   12 Asylum_FED9  0.10
>   15 Asylum_FED9  0.09
>   18 Asylum_FED9  0.07
>   Browse[1]> 
>
> I need to add a column indicating rank within each factor group, which I
> currently accomplish like so:
>
>   wc$rank <- 0
>   for(report in as.character(unique(wc$report))) {
>     wc[wc$report==report,]$rank <- 1:sum(wc$report==report)
>   }
>
> I have to wonder whether there's a better way, something that gets rid of
> the for() loop using tapply() or by() or similar.  But I haven't come up
> with anything.
>
> I've tried these:
>
>   by(wc, wc$report, FUN=function(pr){pr$rank <- 1:nrow(pr)})
>
>   by(wc, wc$report, FUN=function(pr){wc[wc$report %in% pr$report,]$rank <-
> 1:nrow(pr)})
>
> But in both cases the effect of the assignment is lost, there's no $rank
> column generated for wc.
>
> Any suggestions?
>   
There's a little known and somewhat unfortunately named function called 
ave() which does just that sort of thing.

 > ave(wc$score, wc$report, FUN=rank)
 [1] 10.0  9.0  8.0  8.0  7.0  7.0  5.5  5.5  6.0  5.0  4.0  3.5  3.5  
2.0  1.0
[16]  3.0  2.0  1.0