R: Constructor for Profiler Objects from nls Objects

profiler.nls {nls}

R Documentation

Constructor for Profiler Objects from nls Objects

Description

Create a profiler object for the model object fitted of class nls.

Usage

profiler.nls(fitted)

Arguments

fitted

the original fitted model object of class nls.

Value

An object of class profiler.nls which is a list with function elements

getFittedModel()

the nlsModel object corresponding to fitted

getFittedPars()

See documentation for profiler

setDefault(varying, params)

See documentation for profiler

getProfile(varying, params)

In the returned list, fstat is the ratio of change in sum-of-squares and the residual standard error.

For other details, see documentation for profiler

WARNING

When using setDefault and getProfile together, the internal state of the fitted model may get changed. So after completing the profiling for a parameter, the internal states should be restored by a call to setDefault without any arguments. For example see below or the source for profile.nls.

Author(s)

Douglas M. Bates and Saikat DebRoy

References

Bates, D.M. and Watts, D.G. (1988), Nonlinear Regression Analysis and Its Applications, Wiley

Examples

library( nls )
data( BOD )
# obtain the fitted object
fm1 <- nls(demand ~ SSasympOrig( Time, A, lrc ), data = BOD)
# get the profile for the fitted model
prof1 <- profiler( fm1 )
# profile with A fixed at 16.0
prof1$getProfile(c(F,T), 16.0)
# vary lrc
prof1$setDefault(varying = c(F, T))
# fix A at 14.0 and starting estimate of lrc at -0.2
prof1$setDefault(params = c(14.0, -0.2))
# and get the profile
prof1$getProfile()
# finally, set defaults back to original estimates
prof1$setDefault()