reg_control: Tuning parameters for robust regression

In robmed: (Robust) Mediation Analysis

Tuning parameters for robust regression

Description

Obtain a list with tuning paramters for the robust MM-estimator of regression from lmrob() or median regression from rq().

Usage

reg_control(efficiency = 0.85, max_iterations = 200, tol = 1e-07, seed = NULL)

MM_reg_control(
  efficiency = 0.85,
  max_iterations = 200,
  tol = 1e-07,
  seed = NULL
)

median_reg_control(algorithm = "br")

Arguments

efficiency	a numeric value giving the desired efficiency (defaults to 0.85 for 85% efficiency).
max_iterations	an integer giving the maximum number of iterations in various parts of the algorithm.
tol	a small positive numeric value to be used to determine convergence in various parts of the algorithm.
seed	optional initial seed for the random number generator (see .Random.seed).
algorithm	a character string specifying the algorithm for computing the median regression fit. See argument method in rq() for options. If you experience an infinite loop with the default algorithm ("br"), you may want to try "fn".

Details

Prior to version 1.2.0, the MM-estimator of regression was the only type of robust regression in robmed that supported control parameters. Starting with version 1.2.0, control parameters can also be passed to median regression, specifically the type of algorithm to be used. Function reg_control() is an alias for MM_reg_control() for backwards compatibility, but it is now recommended to use MM_reg_control() when performing MM-regression and median_reg_control() when performing median regression.

Value

For MM_reg_control() and reg_control(), a list of control parameters for the MM-estimator of regression as returned by lmrob.control().

For median_reg_control(), a list of control parameters for median regression.

Note

MM_reg_control() and its alias reg_control() is a simplified wrapper function for lmrob.control(), as the latter requires detailed knowledge of the algorithm for the MM-estimator of regression. Currently only 95%, 90%, 85% (the default) and 80% efficiency are supported. For other values, please specify the corresponding tuning parameters in lmrob.control() directly.

Author(s)

Andreas Alfons

References

Salibian-Barrera, M. and Yohai, V.J. (2006) A Fast Algorithm for S-regression Estimates. Journal of Computational and Graphical Statistics, 15(2), 414–427. doi:10.1198/106186006x113629.

Yohai, V.J. (1987) High Breakdown-Point and High Efficiency Estimates for Regression. The Annals of Statistics, 15(20), 642–656. doi:10.1214/aos/1176350366.

Koenker, R.W. (2005) Quantile Regression. Camebridge University Press.

See Also

lmrob(), lmrob.control()

Examples

data("BSG2014")

# run fast-and-robust bootstrap test
ctrl <- MM_reg_control(efficiency = 0.95)
boot <- test_mediation(BSG2014,
                       x = "ValueDiversity",
                       y = "TeamCommitment",
                       m = "TaskConflict",
                       level = 0.9,
                       control = ctrl)
summary(boot)

robmed documentation built on April 4, 2025, 1:10 a.m.