thr3ads.net - R devel - Preliminary version of ts package [Jul 1999]

If this information is useful, please help other people find it:
Share via:

Prof Brian D Ripley

1999-Jul-27 07:06 UTC

Preliminary version of ts package

There is now a preliminary version of a time series package in the R-devel
snapshots, and we would welcome feedback on it.  It is based in part on the
packages bats (Martyn Plummer) and tseries (Adrian Trapletti) and in part
on code I had or have written. (Thanks for the contributions, Martyn and
Adrian!) Some of the existing ts code has been changed, for example to plot
multiple time series, so the behaviour after library(ts) may differ from
that without it.

library(help=ts) gives something like the listing below.  It is intended
that ar.burg (univariate case) and preliminary arima fitting and
forecasting functions will be added in the next couple of weeks.  At 
present there is a fairly high degree of S-PLUS compatibility in the
interfaces, but not necessarily in the objects or algorithms.

----------------------

Preliminary time-series package for R 0.65
=========================================
This is a preliminary version of a time-series package; at present
the functionality is somewhat limited in that not all the options are
yet implemented or may be very slow.

Functions in base R:
-------------------

ts		Create a (univariate or multivariate) ts object
[.ts		Subsetting method for ts objects.
as.ts, is.ts	Coercion and membership functions
plot, print	Methods for plot and print

aggregate	Computes summaries (e.g. sum) over disjoint time intervals
diff		Lagged differences of a time series
end		Time of last observation
frequency	Number of observations per unit of time
start		Time of first observation
time		Create time series giving the times of observations
tsp, tsp<-	Get and set time-series attributes
window		Subset to a time window


Functions in package ts:
-----------------------

acf		Autocovariance and autocorrelation function
ar              Wrapper for autoregression estimation functions
ar.yw           Estimate autoregression model by solving Yule-Walker equations
Box.test	Box-Pierce and Ljung-Box tests of independence
ccf		Cross-covariance and cross-correlations for two series
cpgram		Plot cumulative periodogram of univariate time series
cycle		Create time series giving the positions in a cycle 
		   of a time series
deltat		Return time interval between observations
diffinv		Discrete integration, the inverse of diff()
embed           Embedding a time series
filter		Linear filtering on a time series
kernel          Smoothing kernel objects (and Daniell, Fejer and 
                   (modified) Dirichlet kernels)
lag		Compute lagged version of time series
na.contiguous	Find longest contiguous stretch of non-NAs
na.omit.ts	na.omit method for time series
pacf		Partial autocorrelation function
plot.acf        Plot autocorrelation function
plot.spec       Plot spectral density estimate
plot.spec.coherency
plot.spec.phase
PP.test		Phillips-Perron test for unit root
spec.ar	        Estimate spectral density by autoregression
spec.pgram      Estimate spectral density from periodogram
spec.taper      Taper by cosine bell
spectrum        Wrapper for spectral density estimation functions
stl		Seasonal decomposition using loess
toeplitz	Generate Toeplitz matrix
ts.intersect	Bind time series as multivariate ts over the common time base
ts.union	Bind time series as multivariate ts over their total time base

TO DO

ar.burg		maybe
arima fitting and forecasting.


Datasets in base R:
------------------

airmiles        Passenger-Miles on US Airlines 1937-1960
co2             Moana Loa Atmospheric CO2 Concentrations
nhtemp          Yearly Average Temperatures in New Haven CT
presidents      Quarterly Approval Ratings for US Presidents
sunspots        Monthly Mean Relative Sunspot Numbers
uspop		Populations Recorded by the US Census


Datasets in package ts:
----------------------

accdeaths       US accidental deaths 1973-8
beaver1         time series of body temperatures of two beavers
beaver2
deaths          time-series on UK lung deaths 1974-9 from Diggle (1990)
mdeaths, fdeaths  as above, for males and females
drivers         time series on UK road deaths of drivers from Harvey (1989)
finance         daily closing prices of major European stock indices, 1991-1998
lh              dataset on luteinizing hormone from Diggle (1990)
nottem		time-series of temperatures in Nottingham, 1920-1939
sales		sales data with leading indicator
sunspot.year	yearly sunspot data, 1700-1988
sunspot.month	monthly sunspot data, 1749-1997
treering	yearly tree ring data, -6000-1979


-- 
Brian D. Ripley,                  ripley@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 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Paul Gilbert

1999-Jul-27 17:55 UTC

head link

Preliminary version of ts package

Brian

I haven't got far, but here are some initial problems:
> start(1:10)S:  [1] 1 1
ts: 1
> end(1:10)S: [1] 10  1
ts: 10

window.ts  should handle NULL for default start and end values. If it is going
to print a warning then it needs warn=F to suppress the warning. Below is a copy
of my window.ts, but it would require a few minor changes to work without my
tframe library.

Paul
_______

 window.ts <- function (x, start = NULL, end = NULL, warn = T, eps
.Options$ts.eps)
{
    f <- tsp(x)[3]
    if (is.null(start))
        start <- tsp(x)[1]
    if (is.null(end))
        end <- tsp(x)[2]
    if (2 == length(start))
        start <- start[1] + (start[2] - 1)/f
    if (2 == length(end))
        end <- end[1] + (end[2] - 1)/f
    if (start < tsp(x)[1]) {
        start <- tsp(x)[1]
        if (warn)
            warning("Specified start earlier than start of data, start not
changed.")
    }
    if (end > tsp(x)[2]) {
        end <- tsp(x)[2]
        if (warn)
            warning("Specified end later than end of data, end not
changed.")
    }
    leave <- (time(x) >= (start - eps)) & (time(x) <= (end + eps))
    if (is.matrix(x))
        z <- unclass(x)[leave, , drop = F]
    else z <- x[leave]
    classed(ts(z, start = start, end = end, frequency = f), tfclass(x))
}


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Paul Gilbert

1999-Jul-28 17:41 UTC

head link

Preliminary version of ts package

PG>> I think in general this has been agreed to several times, but also
in
PG>> the discussion of Adrian Trapletti tseries library, which seemed to

PG>> initiate the ts library effort, there was general agreement on this

PG>> point. (See r-devel messages July 15 Re: tseries).

BDR>No, it did not! I promised to do this at the Vienna workshop in
March: the
BDR>end of the Oxford academic year precipitated this. And my reading of
that
BDR>correspondence is _not_ that there was agreement to follow all of
S's
BDR>quirks.  R-core does not think it has agreed to that: I had checked,

BDR>and the consensus is to improve.

There seems to be a general plan that was discussed by R-core and not
mentioned on R-devel, even when there were announcements about other
work being done on time series libraries. If S compatibility is no
longer a consideration then perhaps that should be communicated to
R-devel and even more widely. There are a number of users that have a
big investment in this.

BDR>Certainly we do not intend to get the same
BDR>wrong answers as S-PLUS!

I consider wrong answers to be special exceptions. Changing R's default
behavior for start() and end() is a design choice, not a wrong answer. I
think is is quiet reasonable to put these design choices in your own
library, but it should not be included as part of the default.

Paul



-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Prof Brian D Ripley

1999-Jul-28 19:43 UTC

head link

Preliminary version of ts package

On Wed, 28 Jul 1999, Paul Gilbert wrote:

Paul had just sent a public reply to part of a private reply, a practice I
deplore.

In the part he omitted was a challenge to supply a completely S compatible
lm and glm, plus the comment that I have _not_ changed start.ts nor end.ts
in base R.

-- 
Brian D. Ripley,                  ripley@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 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Possibly Parallel Threads

Search for more maybe matching threads

R devel - Jul 1999 - Preliminary version of ts package

Preliminary version of ts package

Preliminary version of ts package

Preliminary version of ts package

Preliminary version of ts package

Possibly Parallel Threads