[R] Multivariate ARMA Model

[Paul Gilbert]

Hagen Schmöller wrote:

> Hi R-Community,
> 
> so far I dealt with univariate processes and used the function "arima" to
> estimate an ARMA(1,1)-model. For multivariate processes there are the
> functions "estVARXar" and "estVARXls" from package "DSE". But how can I
> estimate an VARMA(1,1)-model, 

You should look at estMaxLik and the dse Users' Guide examples on 
specifying models. The Users' Guide gets installed in 
dse1/doc/dse-guide.pdf. (In dse the ARMA class of models includes VARMA, 
VARMAX, ARIMA, VARIMA, VARIMAX, etc.)

or even better determine the orders and
> estimate the parameters?

If you want to do this automatically, I suggest looking at the 
estimation method called bft in dse1. It estimates VAR models at 
different lag lengths, converts them to state space and does a 
reduction, and then compares the different results. The state space 
reduction results in a more parsimonious representation. This often has 
only an ARMA equivalent and not a VAR equivalent so you should not think 
of this as limiting yourself to VAR models, even thought the estimation 
starts with one. In my experience this does about as well as you can 
hope to do automatically.

If you want to do it manually, estimate the different models you want to 
consider and then compare them. The function informationTests in dse1 is 
a good place to start for doing comparisons.
> 
> Much thanks in advance,
> 
> Hagen Schmoeller
> --
> Dipl.-Ing. Hagen K. Schmöller
> Leiter Forschungsgruppe Stromerzeugung und -handel
> Institut für Elektrische Anlagen und Energiewirtschaft, RWTH Aachen
> Schinkelstraße 6, D-52056 Aachen, Germany
> Tel.: +49 (0)241 80-96734
> Fax : +49 (0)241 80-92197
> Hagen.Schmoeller at iaew.rwth-aachen.de
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
>