rrcov.control function

Control Settings for covMcd and ltsReg

Auxiliary function for passing the estimation options as parameters to the estimation functions.

rrcov.control(alpha = 1/2, method = c("covMcd", "covComed", "ltsReg"),
              nsamp = 500, nmini = 300, kmini = 5,
              seed = NULL, tolSolve = 1e-14,
              scalefn = "hrv2012", maxcsteps = 200,
              trace = FALSE,
              wgtFUN = "01.original", beta,
              use.correction = identical(wgtFUN, "01.original"),
              adjust = FALSE)

Arguments

• alpha: This parameter controls the size of the subsets over which the determinant is minimized, i.e., alpha*n observations are used for computing the determinant. Allowed values are between 0.5 and 1 and the default is 0.5.

• method: a string specifying the main function for which rrcov.control() is used. This currently only makes a difference to determine the default for beta.

• nsamp: number of subsets used for initial estimates or "best"

  or "exact". Default is nsamp = 500. If nsamp="best" exhaustive enumeration is done, as far as the number of trials do not exceed 5000. If nsamp="exact"

  exhaustive enumeration will be attempted however many samples are needed. In this case a warning message will be displayed saying that the computation can take a very long time.

• nmini, kmini: for covMcd: For large $n$, the algorithm splits the data into maximally $kmini$ subsets of targetted size nmini. See covMcd for more details.

• seed: initial seed for R's random number generator; see .Random.seed and the description of the seed

  argument in lmrob.control.

• tolSolve: numeric tolerance to be used for inversion (solve) of the covariance matrix in mahalanobis.

• scalefn: (for deterministic covMcd():) a character string or function for computing a robust scale estimate. The current default "hrv2012" uses the recommendation of Hubert et al (2012); see covMcd for more.

• maxcsteps: integer specifying the maximal number of concentration steps for the deterministic MCD.

• trace: logical or integer indicating whether to print intermediate results. Default is trace = FALSE.

• wgtFUN: a character string or function, specifying how the weights for the reweighting step should be computed, see ltsReg, covMcd or covComed, respectively. The default is specified by "01.original", as the resulting weights are 0 or 1. Alternative string specifications need to match names(.wgtFUN.covComed) - which currently is experimental.

• beta: a quantile, experimentally used for some of the prespecified wgtFUNs, see e.g., .wgtFUN.covMcd and .wgtFUN.covComed.

• use.correction: whether to use finite sample correction factors. Defaults to TRUE.

• adjust: (for ltsReg():) whether to perform intercept adjustment at each step. Because this can be quite time consuming, the default is adjust = FALSE.

See Also

For details, see the documentation about ltsReg and covMcd, respectively.

Returns

A list with components, as the parameters passed by the invocation

Author(s)

Valentin Todorov

Examples

data(Animals, package = "MASS")
brain <- Animals[c(1:24, 26:25, 27:28),]
data(hbk)
hbk.x <- data.matrix(hbk[, 1:3])

ctrl <- rrcov.control(alpha=0.75, trace=TRUE)
covMcd(hbk.x,      control = ctrl)
covMcd(log(brain), control = ctrl)