Idioms for a timeseries operation - moving window

Hi Guys,

Need your wisdom on this.

Say I have a time series (in zoo format) like this

x <- zoo(11:21)
x

 1  2  3  4  5  6  7  8  9 10 11
11 12 13 14 15 16 17 18 19 20 21


I want to do a "moving window sampling" of it. The result can either be a matrix or a dataframe like this

my.super.moving.window(x, length=3, by=1)

11 12 13
12 13 14
13 14 15
14 15 16
....
18 19 20
19 20 21

This shouldn't be new. Many many people must have done this before. Any idea what's the best(efficient and elegant) way to do this?

Cheers,

Bo

Ah! Gabor is the man.

BTW I don't have zoo-faq. (I actually googled it out)

This is what I have:

Vignettes in package 'zoo':

zoo-quickref                        zoo Quick Reference (source, pdf)
zoo                                 zoo: An S3 Class and Methods for Indexed
Totally Ordered Observations
                                    (source, pdf)

It's a typical windows xp install of R 2.6.1.

Any idea why?

Cheers,

Bo

Date: Sun, 2 Mar 2008 09:40:09 -0500
From: ggrothendieck at gmail.com
To: bozhou1981 at hotmail.com
Subject: Re: [R] Idioms for a timeseries operation - moving window
CC: r-help at r-project.org

See

?embed

and from zoo see:
?rollapply
?rollmean
?rollmax

There is a function coded in C in the caTools package for speed.


On Sun, Mar 2, 2008 at 9:24 AM, Bo Zhou <bozhou1981 at hotmail.com> wrote:

Hi Guys,

Need your wisdom on this.

Say I have a time series (in zoo format) like this

x <- zoo(11:21)
x

1 2 3 4 5 6 7 8 9 10 11
11 12 13 14 15 16 17 18 19 20 21


I want to do a "moving window sampling" of it. The result can either be

a matrix or a dataframe like this

my.super.moving.window(x, length=3, by=1)

11 12 13
12 13 14
13 14 15
14 15 16
....
18 19 20
19 20 21

This shouldn't be new. Many many people must have done this before. Any

idea what's the best(efficient and elegant) way to do this?

Cheers,

Bo

star power.

[[alternative HTML version deleted]]

I needed that matrix for following calculations. So I don't think roll*
would work for me.

Re: embed

x <- 1:10
embed (x, 3)

     [,1] [,2] [,3]
[1,]    3    2    1
[2,]    4    3    2
[3,]    5    4    3
[4,]    6    5    4
[5,]    7    6    5
[6,]    8    7    6
[7,]    9    8    7
[8,]   10    9    8

x

 [1]  1  2  3  4  5  6  7  8  9 10

The matrix is in revered order. Anyway to put the matrix in the original
order? Ie
1 2 3
2 3 4
...


Cheers,
Bo

Date: Sun, 2 Mar 2008 09:40:09 -0500
From: ggrothendieck at gmail.com
To: bozhou1981 at hotmail.com
Subject: Re: [R] Idioms for a timeseries operation - moving window
CC: r-help at r-project.org

See

?embed

and from zoo see:
?rollapply
?rollmean
?rollmax

There is a function coded in C in the caTools package for speed.


On Sun, Mar 2, 2008 at 9:24 AM, Bo Zhou <bozhou1981 at hotmail.com> wrote:

Hi Guys,

Need your wisdom on this.

Say I have a time series (in zoo format) like this

x <- zoo(11:21)
x

1 2 3 4 5 6 7 8 9 10 11
11 12 13 14 15 16 17 18 19 20 21


I want to do a "moving window sampling" of it. The result can either be

a matrix or a dataframe like this

my.super.moving.window(x, length=3, by=1)

11 12 13
12 13 14
13 14 15
14 15 16
....
18 19 20
19 20 21

This shouldn't be new. Many many people must have done this before. Any

idea what's the best(efficient and elegant) way to do this?

Cheers,

Bo

star power.

[[alternative HTML version deleted]]