Control parameters for gradient search estimation

A list of parameters for controlling the fitting process.

Usage

nlControl(tol_ll = 1e-05, tol_theta = 0.001, check_theta = FALSE,
  step = NULL, beta = 0.5, gamma = 1.25, reset_step = FALSE,
  maxit = 1000, smooth = FALSE, omicron = 0.001, verbose = FALSE)

Arguments

tol_ll: tolerance expressed as relative change of the objective function.
tol_theta: tolerance expressed as relative change of the estimates.
check_theta: logical flag. If TRUE the algorithm performs a check on the change in the estimates in addition to the likelihood.
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 re-setted to the initial value at each iteration.
maxit: maximum number of iterations.
smooth: logical flag. If TRUE the standard loss function is replaced with a smooth approximation.
omicron: small constant for smoothing the loss function when using smooth = TRUE. See details.
verbose: logical flag.

Value

a list of control parameters.

Details

The optimization algorithm is along the lines of the gradient search algorithm (Bottai et al, 2015). If smooth = TRUE, the classical non-differentiable loss function is replaced with a smooth version (Chen and Wei, 2005).

References

Bottai M, Orsini N, Geraci M (2015). A Gradient Search Maximization Algorithm for the Asymmetric Laplace Likelihood, Journal of Statistical Computation and Simulation, 85(10), 1919-1925.

Chen C, Wei Y (2005). Computational issues for quantile regression. Sankhya: The Indian Journal of Statistics, 67(2), 399-417.

Author

Marco Geraci