[R] correlation structure in lme without random effect

[Douglas Bates]

On 10/30/06, Benjamin Tyner <btyner at stat.purdue.edu> wrote:
> I was hoping to fit along the lines of
>
> g<-gl(20,5)
> y<-runif(100)
> fit<-lme(fixed=y~g,correlation=corAR1(0,~1|g))
>
> But I get the error "Incompatible formulas for groups in "random" and
> "correlation""

Use the gls function in the nlme package to fit a model with
correlation but no random effects specification.