gamlss2

Control Parameters

Description

Control parameters for fitting GAMLSS models with gamlss2.

Usage

gamlss2_control(optimizer = RS, trace = TRUE,
  flush = TRUE, light = FALSE, expand = TRUE,
  model = TRUE, x = TRUE, y = TRUE,
  fixed = FALSE, ...)

Arguments

optimizer Function, the optimizer to be used for fitting.

trace Logical, should information be printed while the algorithm is running?

flush Logical, whether to use flush.console to display current output in the console.

light Logical, if set to light = TRUE, no model frame, response, model matrix and other design matrices will be part of the return value.

expand Logical, if fewer formulas are supplied than there are parameters of the distribution, should intercept-only formulas be added automatically?

model Logical, should the model frame be included as a component of the returned object.

x Logical, indicating whether the model matrix should be included as a component of the returned object.

y Logical, should the response be included as a component of the returned object.

fixed Named logical vector indicating which parameters should be fixed during estimation. See gamlss2_start for examples.

… Further control parameters to be included in the return value, for example parameters used by the optimizer function RS.

Details

The set of control parameters can be extended. For example, if a different optimizer is used, newly specified control parameters are passed on to that optimizer automatically.

For the RS and CG optimizers, argument demmler.reinsch controls the use of a weighted Demmler–Reinsch reparameterization when selecting the smoothing parameter of an eligible univariate, single-penalty smooth. The default, “auto”, starts with the direct solver and switches only when the current smoothing-parameter search reaches the edge of its one-decade search window. This indicates that another costly optimizer restart is needed and amortizes the setup of the reparameterization. Set demmler.reinsch = TRUE to force it for eligible updates, or FALSE to disable it. To preserve the numerical trajectory of nonlinear distributional models, the initial outer iteration continues to use the direct solver.

Smooths with fixed smoothing parameters or multiple penalties, including adaptive and tensor-product smooths, always use the direct solver.

Value

A list with the arguments specified.

Examples

library("gamlss2")

f <- dist ~ s(speed) | s(speed)

## estimate model with different step length
## control in the RS algorithm
m1 <- gamlss2(f, data = cars, family = GA, step = 1)
m2 <- gamlss2(f, data = cars, family = GA, step = 0.1)