Hullo
I may have missed something blindingly obvious here. I'm using xts to
handle some timeseries data. I've got daily measurements for 100
years. If I try to reduce the error rate by taking means of each
month, I'm getting what at first sight appears to be conflicting
information. Here's a small subset to show the problem:
A small set of data:
> vv
x
2010-02-01 6.1
2010-02-02 6.1
2010-02-03 6.0
2010-02-04 6.0
2010-02-05 6.0
2010-02-06 6.1
2010-02-07 6.1
2010-02-08 6.1
2010-02-09 6.1
2010-02-10 6.2
Aggregate:
> aggregate (vv, as.yearmon (index (vv)), mean)
Feb 2010 6.08
That's fine. But if I explicitly convert to xts (which the answer
ought to be, so this should be a noop), the values shift back by one
month:
> xts (aggregate (vv, as.yearmon (index (vv)), mean))
x
Jan 2010 6.08
Just to confirm the classes:
> class (aggregate (vv, as.yearmon (index (vv)), mean))
[1] "zoo"
> class (vv)
[1] "xts" "zoo"
And to confirm that as.yearmon is returning the right month:
> as.yearmon (index (vv))
[1] "Feb 2010" "Feb 2010" "Feb 2010" "Feb
2010" "Feb 2010" "Feb 2010"
[7] "Feb 2010" "Feb 2010" "Feb 2010" "Feb
2010"
This run was on a stock Fedora 10 build:
> version
_
platform i386-redhat-linux-gnu
arch i386
os linux-gnu
system i386, linux-gnu
status
major 2
minor 10.0
year 2009
month 10
day 26
svn rev 50208
language R
version.string R version 2.10.0 (2009-10-26)
And from installed.packages ():
xts NA NA "GPL-3"
"2.10.0"
zoo NA NA "GPL-2"
"2.10.0"
Any help gratefully received.
Tim
I find the following even more confusing as I thought that xts was a
subclass of zoo and I'd expected that the conversion would have been
more transparent
> aggregate (vv, as.yearmon(index(vv)), mean)
Feb 2010 6.08
> xts (aggregate (vv, as.yearmon(index(vv)), mean))
x
Jan 2010 6.08
> zoo (aggregate (vv, as.yearmon(index(vv)), mean))
x
Feb 2010 6.08
On 8 Apr 2010, at 15:53, Tim Coote wrote:
> Hullo
> I may have missed something blindingly obvious here. I'm using xts
> to handle some timeseries data. I've got daily measurements for 100
> years. If I try to reduce the error rate by taking means of each
> month, I'm getting what at first sight appears to be conflicting
> information. Here's a small subset to show the problem:
>
> A small set of data:
> > vv
> x
> 2010-02-01 6.1
> 2010-02-02 6.1
> 2010-02-03 6.0
> 2010-02-04 6.0
> 2010-02-05 6.0
> 2010-02-06 6.1
> 2010-02-07 6.1
> 2010-02-08 6.1
> 2010-02-09 6.1
> 2010-02-10 6.2
>
> Aggregate:
> > aggregate (vv, as.yearmon (index (vv)), mean)
>
> Feb 2010 6.08
>
> That's fine. But if I explicitly convert to xts (which the answer
> ought to be, so this should be a noop), the values shift back by one
> month:
> > xts (aggregate (vv, as.yearmon (index (vv)), mean))
> x
> Jan 2010 6.08
>
> Just to confirm the classes:
> > class (aggregate (vv, as.yearmon (index (vv)), mean))
> [1] "zoo"
>
> > class (vv)
> [1] "xts" "zoo"
>
> And to confirm that as.yearmon is returning the right month:
> > as.yearmon (index (vv))
> [1] "Feb 2010" "Feb 2010" "Feb 2010"
"Feb 2010" "Feb 2010" "Feb 2010"
> [7] "Feb 2010" "Feb 2010" "Feb 2010"
"Feb 2010"
>
> This run was on a stock Fedora 10 build:
> > version
> _
> platform i386-redhat-linux-gnu
> arch i386
> os linux-gnu
> system i386, linux-gnu
> status
> major 2
> minor 10.0
> year 2009
> month 10
> day 26
> svn rev 50208
> language R
> version.string R version 2.10.0 (2009-10-26)
>
> And from installed.packages ():
> xts NA NA "GPL-3"
"2.10.0"
> zoo NA NA "GPL-2"
"2.10.0"
>
> Any help gratefully received.
>
> Tim
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
I find the following even more confusing as I thought that xts was a
subclass of zoo and I'd expected that the conversion would have been
more transparent
> aggregate (vv, as.yearmon(index(vv)), mean)
Feb 2010 6.08
> xts (aggregate (vv, as.yearmon(index(vv)), mean))
x
Jan 2010 6.08
> zoo (aggregate (vv, as.yearmon(index(vv)), mean))
x
Feb 2010 6.08
On 8 Apr 2010, at 15:53, Tim Coote wrote:
On 8 Apr 2010, at 15:53, Tim Coote wrote:
> Hullo
> I may have missed something blindingly obvious here. I'm using xts
> to handle some timeseries data. I've got daily measurements for 100
> years. If I try to reduce the error rate by taking means of each
> month, I'm getting what at first sight appears to be conflicting
> information. Here's a small subset to show the problem:
>
> A small set of data:
> > vv
> x
> 2010-02-01 6.1
> 2010-02-02 6.1
> 2010-02-03 6.0
> 2010-02-04 6.0
> 2010-02-05 6.0
> 2010-02-06 6.1
> 2010-02-07 6.1
> 2010-02-08 6.1
> 2010-02-09 6.1
> 2010-02-10 6.2
>
> Aggregate:
> > aggregate (vv, as.yearmon (index (vv)), mean)
>
> Feb 2010 6.08
>
> That's fine. But if I explicitly convert to xts (which the answer
> ought to be, so this should be a noop), the values shift back by one
> month:
> > xts (aggregate (vv, as.yearmon (index (vv)), mean))
> x
> Jan 2010 6.08
>
> Just to confirm the classes:
> > class (aggregate (vv, as.yearmon (index (vv)), mean))
> [1] "zoo"
>
> > class (vv)
> [1] "xts" "zoo"
>
> And to confirm that as.yearmon is returning the right month:
> > as.yearmon (index (vv))
> [1] "Feb 2010" "Feb 2010" "Feb 2010"
"Feb 2010" "Feb 2010" "Feb 2010"
> [7] "Feb 2010" "Feb 2010" "Feb 2010"
"Feb 2010"
>
> This run was on a stock Fedora 10 build:
> > version
> _
> platform i386-redhat-linux-gnu
> arch i386
> os linux-gnu
> system i386, linux-gnu
> status
> major 2
> minor 10.0
> year 2009
> month 10
> day 26
> svn rev 50208
> language R
> version.string R version 2.10.0 (2009-10-26)
>
> And from installed.packages ():
> xts NA NA "GPL-3"
"2.10.0"
> zoo NA NA "GPL-2"
"2.10.0"
>
> Any help gratefully received.
>
> Tim
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
Tim Coote
tim at coote.org
vincit veritas