Non-linear regression.

[Daniel Carrera <dcarrera math toronto edu>]

I have just realized that the convergence of the non-linear regression is
much better than I first thought.

Essentially I was making the wrong plots (the wrong x-range).

The initial values (it seems) for which the program finds the wrong answer
are those for which I would expect it to.


If the initial values are even remotely close, the convergence is very good.

It would still be good to help the user pick decent initial values.


Cheers,
Daniel.