[R-sig-ME] nlme-function: Step halving factor reduced below minimum in PNLS step
bbolker at gmail.com
Mon Oct 1 01:00:02 CEST 2012
Christof Kluß <ckluss at ...> writes:
> can I avoid the error "Step halving factor reduced below minimum in PNLS
> step" by setting a control parameter to a lower value?
I believe it's the minScale parameter in nlmeControl() (see ?nlmeControl).
However, there's generally a fairly narrow window between (on the one hand)
problems that work OK with the default parameters and (on the other hand)
problems that can be made to work by tweaking the parameters -- once you
start having to adjust the control parameters, the odds are reasonably
high that you're going to have lots of trouble fitting the model no
matter what you do ...
You might check out AD Model Builder as an alternative, although there's
quite a learning curve ...
More information about the R-sig-mixed-models