[R] efficient rolling rank

[Charles C. Berry]

On Fri, 16 Apr 2010, zerdna wrote:

>
> Could someone give me an idea on how to do rolling ranking, i.e. rank in the
> moving window of last 100 numbers in a long vector? I tried naive solution
> like
>
> roll.rank<-function(v, len){
>    r<-numeric(length(v)-len+1)
>    for(i in len:length(v))
>        r[i-len+1]<-rank(v[(i-len+1):i])[len]
>    r
>
> }
>
> However, it turns out pretty slow even on my rather able Linux box. For
> example, doing roll.rank(rnorm(50000), 100) takes 5 second, so for typical
> data i operate which is matrices of the size 1000 x 50000 i will need to
> wait 1.5 hours for one calculation. Does someone know a trick to properly do
> it quicker?

Vectorize it with embed:

> x <- rnorm(50000)
> system.time(x.rank <- rowSums(x[ -(1:99) ] >= embed(x,100) ))
    user  system elapsed
   0.295   0.131   0.424
> system.time(x.rank.2 <- roll.rank(x,100))
    user  system elapsed
   6.907   0.033   6.940
> all.equal(x.rank,x.rank.2)
[1] TRUE
>

N.B., if there are ties, you may want to adjust when

 	rowSums(x[ -(1:99) ] == embed(x,100) )

is greater than 1.

If you want much faster, package inline would enable you to write 
something equivalent in C.

HTH,

Chuck

> -- 
> View this message in context: http://n4.nabble.com/efficient-rolling-rank-tp2013535p2013535.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at 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.
>

Charles C. Berry                            (858) 534-2098
                                             Dept of Family/Preventive Medicine
E mailto:cberry at tajo.ucsd.edu	            UC San Diego
http://famprevmed.ucsd.edu/faculty/cberry/  La Jolla, San Diego 92093-0901