On Tue, 3 Apr 2001 Olivier.Renaud@pse.unige.ch wrote:
> S users,
>
> > version
> S-PLUS 2000 Professional Edition Release 3 for Microsoft Windows : 2000
>
>
> When I fit an arima model with regression terms, I would like to test
> the significance of these terms. However the following call gives me the
> coefficients but not standard errors and tests:
>
> > summary (arima.mle(y, model=list(order=c(1,0,0)), xreg=cbind(1,eff)))
> ...
> Coeffficients for regressor(s): cbind(1, eff)
> [1] -0.15789 1.18306
> ...
Right. Do a likelihood ratio test (make sure n.cond is the same),
as you have the conditional likelihood in the object.
>
> Note that gls gives the tests on the regressions terms (slighly
> different values for the coefficients, since estimated with a different
> method). But for different reasons, I would prefer to stay in the arima
> framework.
gls at least does full MLE, and so is preferable. It is often a lot
slower, though.
>
> > summary (gls (y ~ eff, correlation = corAR1())
> ...
> Coefficients:
> Value Std.Error t-value p-value
> (Intercept) -0.172742 0.1019278 -1.694752 0.0909
> eff 1.197824 0.1439301 8.322260 <.0001
> ...
>
> Any tip would be appreciated.
>
> Olivier
> --
> ----------------------------------------------------------------------
> Olivier Renaud mailto:Olivier.Renaud@pse.unige.ch
> Faculty of Psychology and Education Tel: +41 22 705 9104
> University of Geneva - 40, Bd du Pont d'Arve - CH-1211 Geneva 4
> http://www.unige.ch/fapse/PSY/persons/prof-ana/
> ---------------------------------------------------------------------
> This message was distributed by s-news@lists.biostat.wustl.edu. To
> unsubscribe send e-mail to s-news-request@lists.biostat.wustl.edu with
> the BODY of the message: unsubscribe s-news
>
--
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
|