[R-sig-ME] algal nonlinear mixed effects problem

[Ben Bolker]

Ben Bolker <bbolker at ...> writes:

> 
> 
>   Inspired by a recent question of Diego Pujoni's on the list
> ( http://article.gmane.org/gmane.comp.lang.r.lme4.devel/9363/match= ),
> I started working on the problem, with two realizations:
> 
>  1. the problem really needs a nonlinear fit: see
> http://rpubs.com/bbolker/3363
> 
>  2.  I have no idea how to do an nlmer fit with a non-trivial
> fixed-effects model (e.g. in nlme we could do
> 
> nlmefit1 <- nlme(model = X ~ SSfpl(Day, asymp.L, asymp.R,
>                                    xmid, scale),
>    fixed = list(asymp.R ~ Group, xmid + scale + asymp.L ~ 1),
>    random = asymp.R ~ 1 | Individual, ...)
> 
>   This was raised earlier on stack overflow:
> 
>
http://stackoverflow.com/questions/11056625/how-to-add-fixed-effect-to-four-parameter-logistic-model-in-nlmer
> 

  Reinhold Kliegl e-mailed off-list to say that he was able to do 
it / had to do it by hacking up an objective function that accounts
for the grouping (i.e., more or less doing it by hand).  I have
implemented this and updated http://rpubs.com/bbolker/3363 , which
also shows a solution with AD Model Builder; it turns out that at
least the development version of nlmer is a little too fragile to
fit the model I wanted to fit (I haven't tested with the stable
version).