olmm-control: Control parameters for olmm.

Description

Various parameters that control aspects for olmm.

Usage

olmm_control(fit = c("nlminb", "ucminf", "optim"), 
             doFit = TRUE, numGrad = FALSE, 
             numHess = numGrad, nGHQ = 7L,
             start = NULL, restricted = NULL, verbose = FALSE, ...)

Arguments

fit

character string. The name of the function to be used for the optimization.

doFit

logical scalar. When FALSE an unfitted olmm object is returned. See details.

numGrad

logical scalar indicating whether the score function should be retrieved numerically.

numHess

logical scalar. Indicates whether the Hess matrix for the variance-covariance matrix should be estimated numerically, which is an approximation of the observed Fisher information. Must be TRUE if numGrad is TRUE

nGHQ

a positive integer specifying the number of quadrature points for the approximation of the marginal Likelihood by numerical integration.

start

a named numeric vector of initial values for the parameters. The parameter must be named in exactly in the way as they appear when the model is fitted.

restricted

a character vector of names of coefficients to be restricted to the initial values. The argument is ignored in case of adjacent category models.

verbose

logical scalar. If TRUE verbose output is generated during the optimization of the parameter estimates.

...

further arguments to be passed to fit.

Value

A list of class olmm_control containing the control parameters.

Details

Initial values may decrease the computation time and avoid divergence. The start argument accepts a vector with named elements according to the column names of the model.matrix. At the time being, initial values for adjacent-categories models must be transformed into the baseline-category model form.

Notice that an additional argument control, e.g., control = list(trace = 1), can be passed access control parameters of the optimizers. For arguments, see ucminf, nlminb or optim.

Examples

Run this code

olmm_control(doFit = FALSE)