R help - Apr 2004 - daily time series in R

I have some daily data and would like to apply some of the time series functions
to it.

I have read various notes in the help archive on this, the latest I found
suggested
that I need to use the irts class (Irregularly spaced time series) for daily
data
since a year does not divide into an integer number of days.

I see why I would have to do that if I have gaps (e.g. only data on weekdays)
but do I still have to do that even if I have no gaps, that is if I have every 
data for every day of the year?  In the latter case, the data is evenly spaced,
it;s just that a year is not an integer multiple of this spacing.

I have the same question for weekly data
(which I might create from daily data sampled only on work or weekdays)

Here again, the data is sampled at regular intervals, but a year is not 
an integer multiple of the distance between samples.

Can you give some suggestions for dealing with this?
Should I first create an irts object and then create a ts from that?

If that is the suggestion, then when I get a ts object I have noticed
that in the plot the time axis has year and a decimal fraction of year.
What's a good way to get more standard time increments on the X axis,
such as months or week start days?

Thanks greatly,

- john muller

On Fri, 30 Apr 2004, John Muller wrote:
> I have some daily data and would like to apply some of the time series
> functions to it.
> 
> I have read various notes in the help archive on this, the latest I found
suggested
> that I need to use the irts class (Irregularly spaced time series) for
daily data
> since a year does not divide into an integer number of days.
> 
> I see why I would have to do that if I have gaps (e.g. only data on
weekdays)
> but do I still have to do that even if I have no gaps, that is if I have
every
> data for every day of the year?  In the latter case, the data is evenly
spaced,
> it;s just that a year is not an integer multiple of this spacing.
You can have daily or weekly data as a "ts" object, provided there are
no
gaps.  You can set the frequency to one if it suits you: the frequency 
only affects the labelling (including of frequencies in spectral 
densities).

The `time series functions' only apply to objects of class "ts",
so you do
need to use that.
> I have the same question for weekly data
> (which I might create from daily data sampled only on work or weekdays)
> 
> Here again, the data is sampled at regular intervals, but a year is not 
> an integer multiple of the distance between samples.
> 
> Can you give some suggestions for dealing with this?
> Should I first create an irts object and then create a ts from that?
> 
> If that is the suggestion, then when I get a ts object I have noticed
> that in the plot the time axis has year and a decimal fraction of year.
> What's a good way to get more standard time increments on the X axis,
> such as months or week start days?
Add an axis yourself: see ?axis.POSIXct.  R is programmable, and not 
everyone's wishes are already programmed in.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595