[19133] in s-news-athena
Re: [S] Akaike's Information Criterion (AIC)
daemon@ATHENA.MIT.EDU (Prof Brian D Ripley)
Tue Sep 7 16:59:04 1999
Date: Tue, 7 Sep 1999 21:54:04 +0100 (BST)
From: Prof Brian D Ripley <ripley@stats.ox.ac.uk>
To: Kin Cheung <kin_cheung@alkermes.com>
Cc: "'s-news@wubios.wustl.edu'" <s-news@wubios.wustl.edu>
In-Reply-To: <78E833D774B4D21190A30008C78C717E433310@exch1-ma.alkermes.com>
Message-Id: <Pine.GSO.4.05.9909072148550.4862-100000@auk.stats>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
On Tue, 7 Sep 1999, Kin Cheung wrote:
>
> Hi there,
>
> Can someone kindly tell me how to use Akaike's Information Criterion
> to select the order of a time series. I have simulated two time series,
> one is an autoregressive process and the other one is the general ARIMA.
>
> I tried the following, but it didn't seem to work.
>
> a1 <-arima.fracdiff.sim(model=list(d=0.3,ar=c(0.8,-0.3),ma=0),n=1024)
> AIC(a1)
>
> Error in UseMethod("AIC"): No specific or default method found
What S-PLUS system are you using? AIC seems to be defined in the nlme 3.x
library, but only for its purposes.
However, here you have done a simulation: AIC refers to a model
and maximum likelihood estimation. If you use arima.fracdiff, you
get a MLE and a log-likelihood (up to positive constant) that you cna use
to compute AIC to compare arima.fracdiff fits (and nothing else unless
you know the constant).
In case you wanted to do this for ordinary arimna.mle fits, you can't,
as that does not do maximum likelihood fitting. ar() purports to use
AIC to choose order, but again does not as it does not do MLE (and in fact
does not compute the correct log-likelihood as a result).
--
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
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
