R help - Oct 2006 - Up- or downsampling time series in R

Hi
 
I have data that is sampled (in time) with a certain frequency and I would
like to express this time series as a time series of a higher (or lower)
frequency with the newly added time points being filled in with NA, 0, or
perhaps interpolated. My data might be regularly or irregularly spaced. For
example, I might have quarterly data that I would like to handle as a
monthly time series with NAs filled in for the missing months.
 
RSiteSearch("upsample") gave one link to a function in the
"waveslim"
package that I'm not familiar with. It seems to me that this would be a
fairly common time series task and thus am hoping to find something in the
more common time series packages/classes such as ts, zoo, tseries, etc...
 
I will now give some example code.
 
If I am "lucky" enough that my data is irregularly spaced, then a
combination of zoo and ts already accomplishes this task.
 
> require(zoo)
[1] TRUE
> dt <- sample(c(1,3,9), 20, replace=TRUE)
> t <- zoo(dt, as.yearmon(Sys.Date()) + cumsum(dt)/12)
> t
Jan 2007 Feb 2007 Nov 2007 Feb 2008 Nov 2008 Dec 2008 Mar 2009 Apr 2009 Jul
2009 Aug 2009 
       3        1        9        3        9        1        3        1
3        1 
Nov 2009 Feb 2010 Nov 2010 Aug 2011 May 2012 Jun 2012 Jul 2012 Oct 2012 Jul
2013 Aug 2013 
       3        3        9        9        9        1        1        3
9        1 
> as.ts(t)
     Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
2007   3   1  NA  NA  NA  NA  NA  NA  NA  NA   9  NA
2008  NA   3  NA  NA  NA  NA  NA  NA  NA  NA   9   1
2009  NA  NA   3   1  NA  NA   3   1  NA  NA   3  NA
2010  NA   3  NA  NA  NA  NA  NA  NA  NA  NA   9  NA
2011  NA  NA  NA  NA  NA  NA  NA   9  NA  NA  NA  NA
2012  NA  NA  NA  NA   9   1   1  NA  NA   3  NA  NA
2013  NA  NA  NA  NA  NA  NA   9   1                
> plot(t)
 
However if the data happens to be regularly spaced, upsampling it isn't
quite as straightforward.
 
> t2 <- zoo(sample(1:3, 20, replace=TRUE), as.yearmon(seq(2000, by=0.5,
length=20)))
> t2
Jan 2000 Jul 2000 Jan 2001 Jul 2001 Jan 2002 Jul 2002 Jan 2003 Jul 2003 Jan
2004 Jul 2004 
       3        3        2        2        1        3        1        2
3        3 
Jan 2005 Jul 2005 Jan 2006 Jul 2006 Jan 2007 Jul 2007 Jan 2008 Jul 2008 Jan
2009 Jul 2009 
       2        2        3        3        2        3        3        2
1        3 
> (t2.ts <- as.ts(t2))
Time Series:
Start = c(2000, 1) 
End = c(2009, 2) 
Frequency = 2 
 [1] 3 3 2 2 1 3 1 2 3 3 2 2 3 3 2 3 3 2 1 3
> plot(t2)
>
 
I would expect this to be as simple as changing the frequency attribute of
t2.ts to 12 but I didn't seem to be able to find out how to do this or if it
is possible.
 
So far, the only way around this that I have found is doing it
"manually" in
the following way:
 
> t2.monthly <- zoo(NA, as.yearmon(seq(from=2000, to=2009.5, by=1/12)))
> window(t2.monthly, as.numeric(time(t2)) ) <- as.numeric(t2)            #
can this be done using "[]" indexing?
> t2.monthly
Jan 2000 Feb 2000 Mar 2000 Apr 2000 May 2000 Jun 2000 Jul 2000 Aug 2000 Sep
2000 Oct 2000 
       3       NA       NA       NA       NA       NA        3       NA
NA       NA 
Nov 2000 Dec 2000 Jan 2001 Feb 2001 Mar 2001 Apr 2001 May 2001 Jun 2001 Jul
2001 Aug 2001 
      NA       NA        2       NA       NA       NA       NA       NA
2       NA 
Sep 2001 Oct 2001 Nov 2001 Dec 2001 Jan 2002 Feb 2002 Mar 2002 Apr 2002 May
2002 Jun 2002 
      NA       NA       NA       NA        1       NA       NA       NA
NA       NA 
Jul 2002 Aug 2002 Sep 2002 Oct 2002 Nov 2002 Dec 2002 Jan 2003 Feb 2003 Mar
2003 Apr 2003 
       3       NA       NA       NA       NA       NA        1       NA
NA       NA 
May 2003 Jun 2003 Jul 2003 Aug 2003 Sep 2003 Oct 2003 Nov 2003 Dec 2003 Jan
2004 Feb 2004 
      NA       NA        2       NA       NA       NA       NA       NA
3       NA 
Mar 2004 Apr 2004 May 2004 Jun 2004 Jul 2004 Aug 2004 Sep 2004 Oct 2004 Nov
2004 Dec 2004 
      NA       NA       NA       NA        3       NA       NA       NA
NA       NA 
Jan 2005 Feb 2005 Mar 2005 Apr 2005 May 2005 Jun 2005 Jul 2005 Aug 2005 Sep
2005 Oct 2005 
       2       NA       NA       NA       NA       NA        2       NA
NA       NA 
Nov 2005 Dec 2005 Jan 2006 Feb 2006 Mar 2006 Apr 2006 May 2006 Jun 2006 Jul
2006 Aug 2006 
      NA       NA        3       NA       NA       NA       NA       NA
3       NA 
Sep 2006 Oct 2006 Nov 2006 Dec 2006 Jan 2007 Feb 2007 Mar 2007 Apr 2007 May
2007 Jun 2007 
      NA       NA       NA       NA        2       NA       NA       NA
NA       NA 
Jul 2007 Aug 2007 Sep 2007 Oct 2007 Nov 2007 Dec 2007 Jan 2008 Feb 2008 Mar
2008 Apr 2008 
       3       NA       NA       NA       NA       NA        3       NA
NA       NA 
May 2008 Jun 2008 Jul 2008 Aug 2008 Sep 2008 Oct 2008 Nov 2008 Dec 2008 Jan
2009 Feb 2009 
      NA       NA        2       NA       NA       NA       NA       NA
1       NA 
Mar 2009 Apr 2009 May 2009 Jun 2009 Jul 2009 
      NA       NA       NA       NA        3 
> points(t2.monthly, type="p", col="blue")
> lines(na.locf(t2.monthly), col="blue")        # as an example of
why I
might want to do this.
> 
 
Similarly, it would be nice if one could conveniently downsample a time
series, choosing to keep only the Nth point, or the sum or the average of
the previous N points, etc... I can see how that particular application
could probably be accomplished relatively easily using rapply and a
subsetting operation.  However it might be nice to have a convenient wrapper
for this.
 
Any help would be appreciated.  Thanks in advance.
 
Tobias

********************
Nedbank Limited Reg No 1951/000009/06. The following link displays the names of
the Nedbank Board of Directors and Company Secretary. [
http://www.nedbank.co.za/terms/DirectorsNedbank.htm ]
This email is confidential and is intended for the addressee only. The following
link will take you to Nedbank's legal notice. [
http://www.nedbank.co.za/terms/EmailDisclaimer.htm ]

On Thu, 26 Oct 2006, Brandt, T. (Tobias) wrote:
> I have data that is sampled (in time) with a certain frequency and I would
> like to express this time series as a time series of a higher (or lower)
> frequency with the newly added time points being filled in with NA, 0, or
> perhaps interpolated. My data might be regularly or irregularly spaced. For
> example, I might have quarterly data that I would like to handle as a
> monthly time series with NAs filled in for the missing months.
Both can be done easily with zoo, examples for both are given in the
package vignettes and the man pages of the package.

1. extend to a finer grid (upsampling)
## generate some time series
z <- zoo(sample(1:3, 20, replace=TRUE),
         as.yearmon(seq(2000, by=0.5, length=20)))
## generate emtpy series on finer grid
z2 <- zoo(,seq(start(z), end(z), by = 1/12))
## merge (returns univariate series, by default filled with NAs)
merge(z, z2)

2. aggregate to a coarser grid (downsampling)
## transform to annual data
as.annual <- function(x) floor(as.numeric(x))
## average within years
aggregate(z, as.annual, mean)
## first observation within year
aggregate(z, as.annual, head, 1)

hth,
Z