Nonlinear Fitting Object

DESCRIPTION:
This object is returned by the function ms() to represent the result of fitting a nonlinear model by general minimization.

STRUCTURE:
The object contains the final parameter values, corresponding function and gradient values, and final values for the flags generated internally in the minimization algorithm. If the model was defined in terms of n contributions from n observations, as in the case of minimizing the negative log-likelihood, the function value and derivatives will also be returned on a per-observation form for use in plots, etc.
parameters:
the final values of the parameters in the estimation.
formula:
the formula used for the estimation.
call:
an image of the call to ms(), but with all the arguments explicitly named, so that the data component of the call will always give the data argument, and so on.
value, gradient,:
the final objective function value (one number), and optionally the corresponding gradient vector and
hessian:
matrix of second derivatives, if these were computed. For hessian, only the lower triangle is returned.
pieces, slopes,:
these are the contributions of the N observations to, respectively, the value, the gradients and the
curves:
hessian of the objective function. The first is a vector of length N, the other two are matrices with N rows. The slopes component is only returned if derivatives are computed and the curves component only if second derivatives are computed.
scale:
the scaling vector used by the optimization algorithm.
opt.parameters,:
these are the floating point and integer parameters used and generated by the underlying Fortran
flags:
algorithm. You hope you don't need to know about them, but if you do, see the documentation for the algorithm dmnf in the Port library.
assign:
a list, mapping elements of the parameters component to the named parameters (e.g., as specified by the start argument). Note that the names attribute of parameters gives the individual names; where an element of assign is of length > 1, the individual parameter names will be extended to be unique. The elements of this list are weakly analagous to the terms in a linear model, and the assign components in this sense serve the same function in both cases.
data:
optionally, a copy of the data argument with the final value of the parameters. This is returned if the control in the fit included data=T. The returned data will be a paremetrized data frame if the data in the fit inherited from "data.frame".

SEE ALSO:
nls.object