0

I'm looking to do some correlation work on two times series represented by 2 columns in a datatable. I basically need to be able to run rollapply on one column and be able to access another column in the called function using the current position of rollapply.

Let's consider the following datatable : 

> DT = data.table(y=c(1,3,6), v=1:9)
> DT
   y v
1: 1 1
2: 3 2
3: 6 3
4: 1 4
5: 3 5
6: 6 6
7: 1 7
8: 3 8
9: 6 9

> ft=function(x) if( mean( x[,2,with=FALSE] ) == 4.5) r=c(r,mean(DT[INDEX:(INDEX+3), mean(v)]) )
> rollapply(DT, width=2, function(x) r=mean(x))

The resulting vector r would be 2 5 since the condition mean( x[,2,with=FALSE] ) == 4.5) is true when rollapply gets on rows 2, 5 and 8 (but there aren't 3 rows below the row 8 so you can't get the mean of them).

> r
[1] 2 5

INDEX is the current row from which rollapply is extracting the vector of length 2 from the datable DT to feed it to the function ft.

I don't know if it's even possible to do that since it might be breaking how the whole apply thing works.

Wicelo
  • 2,358
  • 2
  • 28
  • 44
  • You say "this won't work," but I don't get a sense for what "working" would even look like here (since `ft` is quite strange and I don't know where you plan to use it). Could you provide desired output? – Frank Oct 28 '15 at 14:30
  • I did, the desired output is `r` `2 5`, I can rewrite it. I said it won't work because I'm feeding a datable to rollapply which I'm not sure we can do, and also `INDEX` isn't defined I'm searching for a way to get it.. – Wicelo Oct 28 '15 at 14:35
  • As a workaround you could create an "index" column. – natario Oct 28 '15 at 14:48

0 Answers0