lqmmControl: Control parameters for lqmm estimation

In gforge/lqmm_gforge: Linear Quantile Mixed Models

Description

A list of parameters for controlling the fitting process.

Usage

lqmmControl(
  method = "gs",
  LP_tol_ll = 1e-05,
  LP_tol_theta = 1e-05,
  check_theta = FALSE,
  LP_step = NULL,
  beta = 0.5,
  gamma = 1,
  reset_step = FALSE,
  LP_max_iter = 500,
  UP_tol = 1e-04,
  UP_max_iter = 20,
  startQR = FALSE,
  verbose = FALSE
)

Arguments

method	character vector that specifies the estimation method: "gs" for gradient search (default) and "df" for Nelder-Mead.
LP_tol_ll	tolerance expressed as absolute change of the log-likelihood.
LP_tol_theta	tolerance expressed as absolute change of theta
check_theta	logical flag. If TRUE the algorithm performs an additional check on the change in the estimates.
LP_step	step size (default standard deviation of response).
beta	decreasing step factor for line search (0,1).
gamma	nondecreasing step factor for line search (>= 1).
reset_step	logical flag. If TRUE the step size is reset to the initial value at each iteration.
LP_max_iter	maximum number of iterations
UP_tol	tolerance expressed as absolute change of the scale parameter.
UP_max_iter	maximum number of iterations.
startQR	logical flag. If FALSE (default) the least squares estimate of the fixed effects is used as starting value of theta_x and scale. If TRUE the lqm estimate is used.
verbose	logical flag.

Details

LP (lower loop) refers to the estimation of regression coefficients and variance-covariance parameters. UP (upper loop) refers to the estimation of the scale parameter.

Value

a list of control parameters.

Author(s)

Marco Geraci

See Also

lqmm

gforge/lqmm_gforge documentation built on Dec. 20, 2021, 10:42 a.m.