From: tom soyer <tom.soyer_at_gmail.com>

Date: Sun, 13 Jan 2008 08:12:59 -0600

Date: Sun, 13 Jan 2008 08:12:59 -0600

Thanks Richard. I am just trying to understand exactly what is R's arima doing, and I am having a hard time. It seems that xreg is necessary to force arima to include the constant term, but it appears that exactly how this is done is not documented. If a series is not differenced, e.g. AR(1), then does one still need to include xreg? If I take an AR(1) series with known coefficient and use arima to do the fit using arima, i.e., arima(x = ar1, order = c(1, 0, 0)), the estimated coefficient is not as good as a simple lm fit using a one period lag. I am wondering if this should be the expected result? Then, I tried Robert Shumway's sarima, and it gave me a much better coefficient compared to arima (but still a little bit worse than lm). I don't understand why this is happening. I guess I am even more confused...

On 1/12/08, Richard Saba <sabaric_at_charter.net> wrote:

*>
**> Tom
**>
*

> A constant term is not included in the model if any differencing is

*> specified. The xreg= parameter is used to add other explanatory variables to
**> the model. In your case xreg=1:length(x) adds a vector of 1's to the model.
**> Robert Shumway and David Stoffer's website for their "Time Series Analysis
**> an its Applications with R Examples" text has several very helpful documents
**> posted on the site (http://www.stat.pitt.edu/stoffer/tsa2/index.html)
**> specific to time series analysis. The R ISSUES document address your
**> question.
**>
**>
**>
**> Richard
**>
**>
**>
**> >Hi,
**>
**> >I am trying to understand exactly what xreg does in arima. The
**> documentation for xreg says:"xreg Optionally, a vector or matrix of external
**> regressors, which must have >the same number of rows as x." What does this
**> mean with regard to the action of xreg in arima?
**>
**>
**>
**>
**>
**> >Apparently somehow xreg made the following two arima fit equivalent in R:
**>
**>
**> >arima(x, order=c(1,1,1), xreg=1:length(x))
**>
**> >is the same as
**>
**> > arima(diff(x), order=c(1,0,1))
**>
**> >While I understand the latter fit (I think), I am puzzled with regard to
**> the former. Does anyone know what the former is doing to arima, and why it
**> works as it does?
**>
**> >Thanks!
**>
**> --
**>
**> >Tom
**>
**>
**>
**> Richard Saba
**>
**> Department of Economics
**>
**> Auburn University
**>
**> Auburn, AL 36849 USA
**>
**> sabaric_at_auburn.edu
**>
**>
**>
**>
**>
*

-- Tom [[alternative HTML version deleted]] ______________________________________________ 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.Received on Sun 13 Jan 2008 - 14:21:27 GMT

Archive maintained by Robert King, hosted by
the discipline of
statistics at the
University of Newcastle,
Australia.

Archive generated by hypermail 2.2.0, at Sun 13 Jan 2008 - 14:30:06 GMT.

*
Mailing list information is available at https://stat.ethz.ch/mailman/listinfo/r-help.
Please read the posting
guide before posting to the list.
*