Control Parameters for Robust Covariance Estimation

DESCRIPTION:

Allows the user to set values affecting the procedure for robust covariance estimation in covRob.

USAGE:

covRob.control(estim, quan = NULL, ntrial = 500, r = 0.45, alpha = 0.05,
tau = 1e-06, tol = 0.001, maxit = 150, nresamp = "Auto", maxres = "Auto",
random.sample = F, tune = 0.95, prob = 0.99, eps = 0.5)

REQUIRED ARGUMENTS:

estim

the robust estimator used in covRob. The choices are: "mcd" for the Fast MCD algorithm of Rousseeuw and Van Driessen, "donostah" for the Donoho-Stahel projection based estimator, "M" for the constrained M estimator provided by Rocke, "pairwiseQC" for the quadrant correlation based pairwise estimator, and "pairwiseGK" for the Gnanadesikan-Kettenring pairwise estimator. Only the parameters used by estim are returned.

OPTIONAL ARGUMENTS:

MCD Parameters:

quan
the number of observations whose covariance determinant will be minimized in the MCD. The default quan is floor((n+p+1)/2) , where n is the number of observations and p is the number of variables. Any quan between the default and n may be specified. If quan is between 0.5 and 1.0 then it is interpreted as the fraction of observations that should be used.

ntrial
the number of random trial subsamples that are drawn for the MCD covariance estimate.

M Parameters:
r

the fraction of contaminated data. The default value is 0.45.

alpha
the fraction of points receiving zero weight. The default value is 0.05.

maxit
an integer specifying the maximum number of iterations allowed in the computation of the M-estimate. The default value is 150.

tau
numeric value specifying the tolerance used to determine the singularity of the estimated scatter matrix.

tol
numeric value specifying the relative precision of the solution of the M-estimate.

Donoho-Stahel Parameters:
nresamp

the number of resamples required; nresamp may not be reached if too many of the size 'p' subsamples, chosen out of the observed vectors, are in a hyperplane. If nresamp=0 , ALL subsamples are taken. The default nresamp is calculated to ensure a breakdown point of eps with probability prob .

tune
the fraction of the data receiving non-zero weight. The default value is .95.

maxres
the maximum number of resamples performed, including those that are discarded due to linearly dependent subsamples. The default value "Auto" is 2 times nresamp .

prob
used to compute nresamp (when nresamp = "Auto"), prob is the probability of high breakdown. The default value is .99.

eps
used to compute nresamp (when nresamp = "Auto"), eps is the fraction of contamination in the data. The default value is .5.

VALUE:

a list of the control parameters.

SEE ALSO:

EXAMPLES:

covRob.control.list <- covRob.control(estim = "M", r = .2)
covRob(woodmod.dat, estim = "M", control = covRob.control.list)