nlsj.control: Control the Iterations in nlsj
In ArkaB-DS/nlsj: JN radical alternative to port of stats::nls()

Description Usage Arguments Value Author(s) References See Also Examples

View source: R/nlsjcontrol.R

Allow the user to set some characteristics of the nls nonlinear least squares algorithm.

nlsj.control(maxiter = 500, tol = 0.00001, minFactor = 1/1024,
	printEval = FALSE, warnOnly = FALSE, scaleOffset = 0,
    nDcentral = FALSE, watch = FALSE, phi = 1, lamda = 0, 
	offset = 100, laminc = 10, lamdec = 0.4, resmax = 10000, 
	rofftest = TRUE, smallsstest = TRUE,
	derivmeth="numericDeriv", altderivmeth="numericDeriv",
	trace=FALSE)

`maxiter`	A positive integer specifying the maximum number of iterations allowed.
`tol`	A positive numeric value specifying the tolerance level for the relative offset convergence criterion.
`minFactor`	A positive numeric value specifying the minimum step-size factor allowed on any step in the iteration. The increment is calculated with a Gauss-Newton algorithm and successively halved until the residual sum of squares has been decreased or until the step-size factor has been reduced below this limit.
`printEval`	a logical specifying whether the number of evaluations (steps in the gradient direction taken each iteration) is printed.
`warnOnly`	a logical specifying whether `nls()` should return instead of signalling an error in the case of termination before convergence. Termination before convergence happens upon completion of `maxiter` iterations, in the case of a singular gradient, and in the case that the step-size factor is reduced below `minFactor`.
`scaleOffset`	a constant to be added to the denominator of the relative offset convergence criterion calculation to avoid a zero divide in the case where the fit of a model to data is very close. The default value of `0` keeps the legacy behaviour of `nls()`. A value such as `1` seems to work for problems of reasonable scale with very small residuals.
`nDcentral`	only when numerical derivatives are used: `logical` indicating if central differences should be employed, i.e., `numericDeriv(*, central=TRUE)` be used.
`watch`	If TRUE, allows iterations to be interactive. Default = FALSE. NOT currently active.
`phi`	Number to use in a diagonal added to the augmented J' J Marquardt matrix in the Nash-Marquardt stabilization of the Gauss-Newton algorithm. Default = 1
`lamda`	Initial value for the lamda parameter used in Levenberg-Marquardt stabilization (note false spelling of lambda as a historical artifact that often detects copying). When zero, the Gauss-Newton method is used. ?? May need to coordinate with algorithm parameter of nlsj()
`offset`	Number to use in a tolerance-free comparison of two numbers. Default = 100. Items a and b are taken as equal if (a + offset) == (b + offset).
`laminc`	Factor to use to increase the Marquardt lamda parameter. Default = 10
`lamdec`	Factor to use to decrease the Marquardt lamda parameter. Default = 0.4
`resmax`	Maximum number of residual evaluations allowed. Not used in nls(). Default 10000 suggested is likely too big for general use.
`rofftest`	When TRUE (default), use the relative offset convergence criterion, modified if appropriate with scaleOffset to deal with small residuals.
`smallsstest`	TRUE if we check for a very small sum of squares as an indicator for terminating the iterations. nls() only uses the relative offset test, and then only with scaleOffset=0.0. Nevertheless, suggest default of TRUE.
`derivmeth`	The method to computer derivatives for the Jacobian. To mimic nls(), this defaults to "numericDeriv".
`altderivmeth`	An alternate derivative method if "derivmeth" is infeasible, e.g., if derivmeth is "default" (analytic), then the alternative "numericDeriv" would call that routine. In future, there may be other methods.??
`trace`	When TRUE, allows progress information to be printed. ?? NEED to coordinate with argument "trace". Default=FALSE)

A list with components

`maxiter`
`tol`
`minFactor`
`printEval`
`warnOnly`
`scaleOffset`
`nDcentreal`
`watch`
`phi`
`lamda`
`offset`
`laminc`
`lamdec`
`resmax`
`rofftest`
`smallsstest`
`derivmeth`
`altderivmeth`
`trace`

with meanings as explained under ‘Arguments’.

Douglas Bates and Saikat DebRoy; John C. Nash for extensions made in the 2020 Improvements to nls() Google Summer of Code project with Arkajyoti Bhattacharjee.

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

nls