robsvyreg: Internal Function for the Regression GM-Estimator

Description

Internal function to call the robust survey regression GM-estimator; this function is only intended for internal use. The function does not check or validate the arguments. In particular, missing values in the data may make the function crash.

Usage

robsvyreg(x, y, w, k, psi, type, xwgt, var = NULL, verbose = TRUE, ...)
svyreg_control(tol = 1e-5, maxit = 100, k_Inf = 1e6, init = NULL,
               mad_center = TRUE, ...)

Value

[list]

Arguments

x: [numeric matrix] design matrix (NA values not allowed).
y: [numeric vector] dependent variable (NA values not allowed).
w: [numeric vector] weights (no NA's allowed).
k: [double] robustness tuning constant (\(0 < k \leq \infty\)).
psi: [integer] psi-functions: 0: Huber, 1: asymmetric Huber, and 2: Tukey biweight.
type: [integer] type of estimator; 0: M-estimator; 1: Mallows and 2: Schweppe type GM-estimator.
xwgt: [numeric vector] weights for design space used in GM-estimators (default: NULL, NA values not allowed).
var: [numeric vector] heteroscedastic variance (default: NULL).
verbose: [logical] indicating whether additional information is printed to the console (default: TRUE).
tol: [double] numerical tolerance criterion to stop the iterations (default: 1e-05).
maxit: [integer] maximum number of iterations to use (default: 100).
k_Inf: [integer] numerical value that represents Inf (default: 1e+06).
init: either NULL or [numeric vector], if init = NULL the regression estimator is initialized by weighted least squares; otherwise, init can be specified as the estimate (i.e., p-vector) to initialize the iteratively re-weighted least squares method (default: NULL).
mad_center: [logical] if TRUE, the weighted MAD is centered about the (weighted) median, otherwise the weighted MAD is centered about zero (default: TRUE).
...: additional arguments passed to the method (see svyreg_control).

Details

Not documented