[R] Re: linear model with arma errors

[Rob Hyndman]

> Dear all,
> 
> I'm looking for how can I estimate a linear model with ar(ma) errors :
> 
> y(t)=a*X(t)+e(t) with
> P(B)e(t)=Q(B)u(t)
> 
> where u is a white noise and P, Q are some polynomes.
> 
> Could you help me ?


Isn't this what arima in library ts does? Put X(t) into xreg and specify
the order of the arma error via the order argument.

The documentation for arima doesn't actually explain how xreg enters the
model, but if it is like the S-Plus arima.mle() function then it does
what you want. The BIG advantage of the R implementation is that it
gives inference for the regression coefficients.

[Aside to Brian Ripley: Can you please add a few lines to the arima help
file to define the model with a non-null xreg?]

Cheers,
Rob


___________________________________________________
Rob J Hyndman
Associate Professor & Director of Consulting
Department of Econometrics & Business Statistics
Monash University, VIC 3800, Australia.
http://www-personal.buseco.monash.edu.au/~hyndman/