[R-sig-ME] gls and AR>1 [Coefficient matrix not invertible]

[Jon Loehrke]

Hi Mixed Modelers,

	I have  been using gls() to run some AR[2] models and occasionally  
receive the following error:

gls(Y~x1 + x2 + x3 + x4 + x5 + x6, data=na.omit(herrdata),  method  
='ML',corr=corARMA(p = 2, q = 0))

	Error in `coef<-.corARMA`(`*tmp*`, value = c(7.49067599726739,  
-15.2313908862033 :
   		Coefficient matrix not invertible

	I do not understand the problem enough to create a dataset and post  
an example.  However, I have established
a few things:

	if I use method='REML' then the error goes away
	if I reduce the number of parameters, the error goes away
	if I utilize an AR[1] correlation structure, the error goes away

	Would it be possible for somebody to explain this error to me?

Thank you very much, Jon

sessionInfo()
R version 2.8.1 Patched (2009-01-19 r47650)
i386-apple-darwin9.6.0

locale:
en_US.UTF-8/en_US.UTF-8/C/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] grid      stats     graphics  grDevices utils     datasets   
methods   base

other attached packages:
[1] nlme_3.1-90        car_1.2-9          xtable_1.5-4        
MASS_7.2-45        RColorBrewer_1.0-2 lattice_0.17-20     
reshape_0.8.2      plyr_0.1.3


Jon Loehrke
Graduate Research Assistant
Department of Fisheries Oceanography
School for Marine Science and Technology
University of Massachusetts
200 Mill Road, Suite 325
Fairhaven, MA 02719
jloehrke at umassd.edu