[R] nlme gls() error

[Richard Nixon]

Hello

I'm fitting a gls model with a variance-covariance structure and an 
getting an error message I don't understand

I'm using gls() from the nlme library with the structure defined by

correlation = corSymm(form = ~1|Subject), weights = varIdent(form=~1|strata)

I get the error

Error in recalc.corSymm(object[[i]], conLin) :
    NA/NaN/Inf in foreign function call (arg 1)

My dependent variable is highly positively skewed and has with many zero 
value.

Any ideas as to the cause of the error? Could I play around with any of 
the  glsControl values to help out?

Thanks
Richard

-- 
Dr. Richard Nixon, MRC Biostatistics Unit, Cambridge, UK
http://www.mrc-bsu.cam.ac.uk/personal/richard