Control Parameters for Heteroscedastic Binary Response GLMs

Various parameters that control fitting of heteroscedastic binary response models using hetglm.

Usage

hetglm.control(method = "nlminb", maxit = 1000,
  hessian = FALSE, trace = FALSE, start = NULL, ...)

Arguments

method

characters string specifying either that nlminb is used for optimization or the method argument passed to optim (typically, "BFGS" or "L-BFGS-B").

maxit

integer specifying the maximal number of iterations in the optimization.

hessian

logical. Should the numerical Hessian matrix from the optim output be used for estimation of the covariance matrix? The default (and only option for nlminb) is to use the analytical expected information rather than the numerical Hessian.

trace

logical or integer controlling whether tracing information on the progress of the optimization should be produced?

start

an optional vector with starting values for all parameters.

...

arguments passed to the optimizer.

Details

All parameters in hetglm are estimated by maximum likelihood using either nlminb (default) or optim with analytical gradients and (by default) analytical expected information. Further control options can be set in hetglm.control, most of which are simply passed on to the corresponding optimizer.

Starting values can be supplied via start or estimated by glm.fit, using the homoscedastic model. Covariances are derived analytically by default. Alternatively, the numerical Hessian matrix returned by optim can be employed, in case this is used for the optimization itself.

Value

A list with the processed specified arguments.

See also

hetglm