This reply only addresses the NaN in Jacobian matter. I believe it is a
result of getting a perfect fit (0 sum of squares). I have amended the
r-forge version of nlmrt package in routines nlfb and nlxb and did not
get the error running Elizabeth's example. This only answers the
software issue,
1. This is a statistical, not an R issue. So I am keeping this offlist.
2. Without a prior family of models, you should **not** be fitting a
parametric nonlinear model.
3. An interpolating smooth is what is wanted. FIt one (e.g. splines,
kernel smooth, etc.)
4. You are out of your depth statist
Elizabeth,
You should cc rhelp on all correspondence so other readers can follow the
thread of conversation.
Now that I have some data to play with, I see where I went wrong in my
previous e-mail.
First of all, you can't fit a model
CO2 ~ log(a*Time) + b
because log(a*Time) can be rewritten
Elizabeth,
It's difficult to troubleshoot without the data. Could you provide the
output from
dput(FG2)
or if your data set is quite large, perhaps
dput(FG2[1:50, ])
If you want to fit a third parameter to represent the base of the log, you
could use
nls(CO2 ~ log(a*Time) / log(c)
4 matches
Mail list logo
