brquasiControl: Auxiliary function for ['glm()'] fitting using the...

In ikosmidis/brquasi: Bias reduction in quasi likelihood estimation

Auxiliary function for ['glm()'] fitting using the ['brquasiFit()'] method.

Description

Typically only used internally by 'brquasiFit()', but may be used to construct a 'control' argument.

Usage

brquasiControl(
  epsilon = 1e-04,
  maxit = 500,
  trace = FALSE,
  type = c("iRBM", "M", "eRBM", "MPQL_trace"),
  slowit = 1,
  response_adjustment = 0,
  max_step_factor = 12,
  only_beta = TRUE,
  lambda = 1e-10,
  disp_factor = c("n-p", "n")
)

brquasi_control(
  epsilon = 1e-04,
  maxit = 500,
  trace = FALSE,
  type = c("iRBM", "M", "eRBM", "MPQL_trace"),
  slowit = 1,
  response_adjustment = 0,
  max_step_factor = 12,
  only_beta = TRUE,
  lambda = 1e-10,
  disp_factor = c("n-p", "n")
)

Arguments

epsilon	positive convergence tolerance epsilon. Default is '1e-04'.
maxit	integer giving the maximal number of iterations allowed. Default is '500'.
trace	logical indicating if output should be produced for each iteration. Default is 'FALSE'.
type	the type of fitting method to be used. The options are '"M"' (standard M-estimation based on quasi-likelihoods), '"iRBM"' (implicit reduced-bias M-estiamtion: empirically adjusted quasi likelihood equations for mean-bias reduction; default), 'eRBM' (explicit reduced-bias M-estimation: correction of asymptotic mean bias using empirical bias estimates). '"MPQL_trace"' (maximum penalized quasi likelihood estimation; see Details).
slowit	a positive real used as a multiplier for the stepsize. The smaller it is the smaller the steps are. Default is '1'.
response_adjustment	a (small) positive constant or a vector of such. Default is 'NULL'. See Details.
max_step_factor	the maximum number of step halving steps to consider. Default is '12'.
only_beta	Should RBM estimation be used for improving estimation of the regression coefficients only? Default is 'TRUE'.
lambda	a ridge adjustment to be added to the diagonal of the jacobian before inverting it for the computation of the step size in the quasi-Fisher iteration. Default is '1e-10'. See Details.
disp_factor	factor by which to divide the sum of pearson residuals when estimating the dispersion. Defauls is '"n-p"' which corresponds to "number of observations" minus "number of parameters". See Details.

Details

TO ADD: details on iRBM, eRBM, disp_factor

ikosmidis/brquasi documentation built on Jan. 27, 2023, 9:04 p.m.