R help - Jul 2005 - computation time gls()

dear R users

i try to fit a gls model to a rather large dataset with an AR(1) error 
structure:

attach(sf.a1filt)
m1.a.gls <- gls(fluxt~co2+light+vpd+wind,
   correlation = corAR1(0.8))
summary(m1.a.gls)
detach(sf.a1filt)

there are approx. 5000 observations, and the computation seems to take several 
hours, i actually killed the process because i became too impatient. is there 
any way to be more efficient with R? (because really the model will be more 
complex, i.e. more predictors and higher autoregressive order). 

os: linux suse, R version: latest, machine: ibm thinkpad 42T, 1GHz RAM


------------------------------------------------
Sebastian Leuzinger
Institute of Botany, University of Basel
Sch??nbeinstr. 6 CH-4056 Basel
ph    0041 (0) 61 2673511
fax   0041 (0) 61 2673504
email Sebastian.Leuzinger at unibas.ch 
web   http://pages.unibas.ch/botschoen/leuzinger

I find arima() fits such models very much faster.

On Mon, 25 Jul 2005, Sebastian Leuzinger wrote:
> dear R users
>
> i try to fit a gls model to a rather large dataset with an AR(1) error
> structure:
>
> attach(sf.a1filt)
> m1.a.gls <- gls(fluxt~co2+light+vpd+wind,
>   correlation = corAR1(0.8))
> summary(m1.a.gls)
> detach(sf.a1filt)
>
> there are approx. 5000 observations, and the computation seems to take
several
> hours, i actually killed the process because i became too impatient. is
there
> any way to be more efficient with R? (because really the model will be more
> complex, i.e. more predictors and higher autoregressive order).
-- 
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 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595