apply a function to a rolling subset of a vector

On Wed, 2 Mar 2005 17:22:43 -0500, "Whit Armstrong"
<whit at twinfieldscapital.com> wrote :

Does anyone know an easy way to calculate the rolling 20 period average
or sum of a vector?

For instance:
x <- rnorm(1000)

y <- apply.subset(x,20,fun="sum")

The first element of y would contain the sum of elements 1 to 20, the
second element of y 
would contain the sum of elements 2:21, and so on.

I thought I had seen this on the list a year or so ago, but I couldn't
find anything in the archives.

I don't know of a general purpose function, but filter() (in the stats
package) can do the example you give, or any other linear filter.

e.g.

x <- rnorm(1000)
y <- filter(x, rep(1,20))

puts 20 element sums into y.  The vector ends up the same length as x,
with NAs at the beginning and end (by default).

Duncan Murdoch

Does anyone know an easy way to calculate the rolling 20 period average
or sum of a vector?

For instance:
x <- rnorm(1000)

y <- apply.subset(x,20,fun="sum")

The first element of y would contain the sum of elements 1 to 20, the
second element of y 
would contain the sum of elements 2:21, and so on.

I thought I had seen this on the list a year or so ago, but I couldn't
find anything in the archives.

You can use the running() function in the gtools package, which is in
the gregmisc bundle:

x <- rnorm(1000)

running(x, fun = sum, width = 20)

1:20          2:21          3:22          4:23          5:24
 -2.009684610  -2.205737077  -1.410810606  -2.226661837  -1.684604289
         6:25          7:26          8:27          9:28         10:29
 -4.492008605  -3.816273719  -5.348364598  -6.444591766  -5.263013812
        11:30         12:31         13:32         14:33         15:34
 -4.609829115  -5.935537291  -6.909232329  -4.881021777  -5.803659103
...

See ?running for more information, after installing gregmisc from CRAN. 

HTH,

Marc Schwartz

Whit Armstrong <whit <at> twinfieldscapital.com> writes:

: 
: Does anyone know an easy way to calculate the rolling 20 period average
: or sum of a vector?
: 
: For instance:
: x <- rnorm(1000)
: 
: y <- apply.subset(x,20,fun="sum")
: 
: The first element of y would contain the sum of elements 1 to 20, the
: second element of y 
: would contain the sum of elements 2:21, and so on.
: 
: I thought I had seen this on the list a year or so ago, but I couldn't
: find anything in the archives.
: 

Look at ?filter .  Also ?embed and gtools::running .  filter is the
fastest.