gsm.control: Defaults for the gsm call
In mclements/rstpm2: Smooth Survival Models, Including Generalized Survival Models

gsm.control

R Documentation

Defaults for the gsm call

Description

Set useful default and allow changes for the gsm call. This is meant to make the gsm call simpler.

Usage

gsm.control(parscale = 1, maxit = 300, optimiser = c("BFGS", "NelderMead"), trace = 0,
            nodes = 9, adaptive = TRUE, kappa.init = 1, maxkappa = 1000,
            suppressWarnings.coxph.frailty = TRUE, robust_initial = FALSE, bhazinit = 0.1,
            eps.init = 1e-5, use.gr = TRUE, penalty = c("logH", "h"), outer_optim = 1,
            reltol.search = 1e-10, reltol.final = 1e-10, reltol.outer = 1e-05,
            criterion = c("GCV", "BIC"))

Arguments

`parscale`	numeric vector or scalar for the scaling of the parameter values; default 1
`maxit`	integer for the maximum number of iterations for the optimisation process
`optimiser`	which optimiser to use for the outer optimisation
`trace`	integer indicating the trace level for each optimiser
`nodes`	number of quadrature nodes
`adaptive`	logical for whether to use adaptive or non-adaptive quadrature, Default=TRUE.
`kappa.init`	initial value for the quadratic penalty for inequality constraints
`eps.init`	initial value for epsilon
`maxkappa`	double float value for the maximum value of the weight used in the constraint.
`suppressWarnings.coxph.frailty`	logical
`robust_initial`	Not currently documented.
`bhazinit`	Not currently documented.
`use.gr`	Logical for whether to use gradients.
`penalty`	Not currently documented.
`outer_optim`	Not currently documented.
`reltol.search`	Relative tolerance. Not currently documented.
`reltol.final`	Relative tolerance. Not currently documented.
`reltol.outer`	Relative tolerance. Not currently documented.
`criterion`	Not currently documented.