Hello everyone,
Since my question is quite theorical, I am not sure whether it is the right
place to ask, but anyway...
I am working on time series and I looked at some way to fit my data through
arima models.
Since these data are updated frequently, I was looking at a way to update
the model "on line" (to get a kind of recursive estimation)
So the next step was to express the arima models as state-space
(structural) models.
The idea was to use the recursive formulaes of a Kalman Filter, in order to
get an estimation of the kind of the recursive least square.
But it seems to me that the estimation of these structural models requires
a likelihood maximization which is not recursive.
So my question is :
In a structural model, can the likelihood maximization be done recursively
?
Upon what I read in 2 first articles of the 2/2 issue of R News, I don't
think it is done this way in R.
Thanks in advance,
Julien
----------------
Les donnees et renseignements contenus dans ce message sont personnels,
confidentiels et secrets. Ce message est adresse a l'individu ou
l'entite dont les coordonnees figurent ci-dessus. Si vous n'etes pas le
bon destinataire, nous vous demandons de ne pas lire, copier, utiliser ou
divulguer cette communication. Nous vous prions de notifier cette erreur a
l'expediteur et d'effacer immediatement cette communication de votre
systeme.
The information contained in this message is privileged, confidential, and
protected from disclosure. This message is intended for the individual or entity
adressed herein. If you are not the intended recipient, please do not read,
copy, use or disclose this communication to others ;also please notify the
sender by replying to this message, and then delete it from your system.
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help 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-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
On Mon, 5 Aug 2002 julien.ruiz at airfrance.fr wrote:> Since my question is quite theorical, I am not sure whether it is the right > place to ask, but anyway... > I am working on time series and I looked at some way to fit my data through > arima models. > Since these data are updated frequently, I was looking at a way to update > the model "on line" (to get a kind of recursive estimation) > So the next step was to express the arima models as state-space > (structural) models. > The idea was to use the recursive formulaes of a Kalman Filter, in order to > get an estimation of the kind of the recursive least square. > But it seems to me that the estimation of these structural models requires > a likelihood maximization which is not recursive. > > So my question is : > In a structural model, can the likelihood maximization be done recursively > ? > > Upon what I read in 2 first articles of the 2/2 issue of R News, I don't > think it is done this way in R.1) ARIMA fitting *is* done via state-space models, but structural models are something different. 2) You can't (in general) do ML estimation of the parameters of a state-space model recursively. Nor is that what recursive least squares estimates. For more details, see the references in the article you mention, especially the Durbin & Koopman book. -- 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 272860 (secr) Oxford OX1 3TG, UK Fax: +44 1865 272595 -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help 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-help-request at stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
Hello Julien, as I understand you correctly, you want to perform pseudo-ex ante forecasts. In this case, you can place the data extraction (i.e. length of the sample period) and the program code for estimation into a for-loop for pseudo ex ante forecasting. Therefore estimating your model recursively over a time span in the past and writing each time the n-step ahead forecasts into another object. The latest estimation is then carried out until: today's period - forecast span. Bernhard -----Original Message----- From: ripley at stats.ox.ac.uk [mailto:ripley at stats.ox.ac.uk] Sent: 05 August 2002 11:36 To: julien.ruiz at airfrance.fr Cc: r-help at stat.math.ethz.ch Subject: Re: [R] Structural TS and recursive estimation On Mon, 5 Aug 2002 julien.ruiz at airfrance.fr wrote:> Since my question is quite theorical, I am not sure whether it is theright> place to ask, but anyway... > I am working on time series and I looked at some way to fit my datathrough> arima models. > Since these data are updated frequently, I was looking at a way to update > the model "on line" (to get a kind of recursive estimation) > So the next step was to express the arima models as state-space > (structural) models. > The idea was to use the recursive formulaes of a Kalman Filter, in orderto> get an estimation of the kind of the recursive least square. > But it seems to me that the estimation of these structural models requires > a likelihood maximization which is not recursive. > > So my question is : > In a structural model, can the likelihood maximization be done recursively > ? > > Upon what I read in 2 first articles of the 2/2 issue of R News, I don't > think it is done this way in R.1) ARIMA fitting *is* done via state-space models, but structural models are something different. 2) You can't (in general) do ML estimation of the parameters of a state-space model recursively. Nor is that what recursive least squares estimates. For more details, see the references in the article you mention, especially the Durbin & Koopman book. -- 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 272860 (secr) Oxford OX1 3TG, UK Fax: +44 1865 272595 -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-. -.- r-help 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-help-request at stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._. _._ If you have received this e-mail in error or wish to read our e-mail disclaimer statement and monitoring policy, please refer to http://www.drkw.com/disc/email/ or contact the sender -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help 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-help-request at stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._