Provide a summary output (but involving some serious computations!) of
an object of class nlsr from nlxb or nlfb from package
nlsr.
# S3 method for nlsr
summary(object, ...)returns an invisible copy of the nlsr object.
An object of class 'nlsr'
Currently ignored.
John C Nash <nashjc@uottawa.ca>
summary.nlsr performs a summary method for an object of class 'nlsr' that
has been created by a routine such as nlfb or nlxb for nonlinear
least squares problems.
Issue: When there are bounded parameters, nls returns a Standard Error for each of
the parameters. However, this summary does NOT have a Jacobian value (it is set to 0)
for columns where a parameter is masked or at (or very close to) a bound. See the
R code for the determination of whether we are at a bound. In this case,
users may wish to look in the inst/dev-codes directory of this package,
where there is a script seboundsnlsrx.R that computes the nls()
standard errors for comparison on a simple problem.
Issue: The printsum() of this object includes the singular values of the Jacobian.
These are displayed, one per coefficient row, with the coefficients. However, the
Jacobian singular values do NOT have a direct correspondence to the coefficients
on whose display row they appear. It simply happens that there are as many Jacobian
singular values as coefficients, and this is a convenient place to display them.
The same issue applies to the gradient components.