[R-sig-ME] error from using variance function on Theoph data

[Michael Li]

Hi, I was trying out nlme function.  For some reason, it did work for
me when I tried to use the power variance function,
even though this is a well known example (and I copied the code from
one of J Pinheiro's presentations).

What could be wrong? I'm using R 2.9.0.

Thanks

Michael

> library (nlme)
> m1 = nlme (conc ~ SSfol (Dose, Time, lKe, lKa, lCl), Theoph,
+ fixed = lKe + lKa + lCl ~ 1,
+ start = list (fixed = c(-2.5, 0.4, -3)),
+ random = pdDiag (lKa + lCl ~ 1))
> m1
Nonlinear mixed-effects model fit by maximum likelihood
  Model: conc ~ SSfol(Dose, Time, lKe, lKa, lCl)
  Data: Theoph
  Log-likelihood: -177.0215
  Fixed: lKe + lKa + lCl ~ 1
       lKe        lKa        lCl
-2.4547019  0.4657267 -3.2272219

Random effects:
 Formula: list(lKa ~ 1, lCl ~ 1)
 Level: Subject
 Structure: Diagonal
              lKa       lCl  Residual
StdDev: 0.6435864 0.1669279 0.7092533

Number of Observations: 132
Number of Groups: 12
> update (m1, weights = varConstPower (power = 0.1))
Error in eigen(val) : infinite or missing values in 'x'