[R] inverse cumsum

[Wacek Kusnierczyk]

Gavin Simpson wrote:
>
>> rc = nrow(m)
>> cumsums = apply(m[rc:1,], 2, cumsum)[rc:1,]
>>     
>
> Alternatively, just use rev() instead of doing it by hand:
>
>
>   
>> cs2 <- apply(dat, 2, function(x) {rev(cumsum(rev(x)))})
>>     
yes, that's more elegant, but somewhat less efficient:

m = matrix(1:100000, 100, 100)
system.time(for (i in 1:1000) {rc=nrow(m); apply(m[rc:1,], 2,
cumsum)[rc:1,]})
#    user  system elapsed
#  1.981   0.000   1.983
system.time(for (i in 1:1000) apply(m, 2, function(x) rev(cumsum(rev(x))))
#   user  system elapsed
#  4.816   0.016   4.833
system.time({revcumsum=function(x) rev(cumsum(rev(x))); for (i in
1:1000) apply(m, 2, revcumsum)})
#   user  system elapsed
#  4.824   0.036   4.866

vQ