Various parameters that control fitting of generalized linear models
with extra parameters using glmx.
glmx.control(profile = TRUE, nuisance = FALSE,
start = NULL, xstart = NULL, hessian = TRUE, method = "BFGS",
epsilon = 1e-8, maxit = c(500, 25), trace = FALSE,
reltol = .Machine$double.eps^(1/1.2), ...)A list with the arguments specified.
logical. Should the extra parameters be optimized via profile likelihood (or via the full likelihood of all parameters)?
logical. Should the extra parameters be treated as nuisance parameters (i.e., suppressed in subsequent output)?
an optional vector with starting values for the GLM coefficients.
an optional vector with starting values for the extra parameter(s). Must be supplied if there is more than one extra parameter.
logical or character. Should the hessian be computed
to estimate the covariance matrix? If character, hessian can
be either "none", "optim" or "numDeriv". The
default is the hessian from optim but alternatively
hessian from the numDeriv package can be used.
characters string specifying the method argument
passed to optim.
numeric convergance tolerance passed to glm.control.
integer specifying the maxit argument (maximal number
of iterations) passed to optim and glm.control.
Can also be a vector of length 2.
logical or integer controlling whether tracing information on
the progress of the optimization should be produced (passed to
optim, and glm.control).
Can also be a vector of length 2.
arguments passed to optim.
All parameters in glmx are estimated by maximum likelihood
using optim with control options set in glmx.control.
Either the parameters can be found by only optimizing over the extra parameters
(and then using glm.fit to estimate the GLM coefficients),
or alternatively all parameters can be optimized simultaneously.
Covariances are derived numerically using the Hessian matrix returned by
optim.
glmx