[R] NLS estimation: Starting values

[Prof Brian Ripley]

optim() needs starting values too, but *may* be more robust than nls to 
their specification ....

On Fri, 24 Oct 2003, Giovanni Caggiano wrote:

> I am trying to fit to the data a nonlinear function, say z(tau;b) where tau 
> is the independent variable and b is 3x1 vector of parameters. Following 
> previous suggestions from the list, I am trying to find some "optimal" 
> starting values for the three parameters by using the command 'optim' and 
> then feed these values to nls for estimation. Does anybody know hot to use 
> 'optim' for this purpose?

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595