Tuning parameters for multivariate S, MM and GS estimates as used in FRB functions for multivariate regression, PCA and Hotelling tests. Mainly regarding the fast-(G)S algorithm.
Scontrol(nsamp = 500, k = 3, bestr = 5, convTol = 1e-10, maxIt = 50)MMcontrol(bdp = 0.5, eff = 0.95, shapeEff = FALSE, convTol.MM = 1e-07,
maxIt.MM = 50, fastScontrols = Scontrol(...), ...)
GScontrol(nsamp = 100, k = 3, bestr = 5, convTol = 1e-10, maxIt = 50)
A list with the tuning parameters as set by the arguments.
number of random subsamples to be used in the fast-(G)S algorithm
number of initial concentration steps performed on each subsample candidate
number of best candidates to keep for full iteration (i.e. concentration steps until convergence)
relative convergence tolerance for estimates used in (G)S-concentration iteration
maximal number of steps in (G)S-concentration iteration
breakdown point of the MM-estimates; usually equals 0.5
Gaussian efficiency of the MM-estimates; usually set at 0.95
logical; if TRUE, eff is with regard to shape-efficiency,
otherwise location-efficiency
relative convergence tolerance for estimates used in MM-iteration
maximal number of steps in MM-iteration
the tuning parameters of the initial S-estimate
allows for any individual parameter from Scontrol to be set directly
Gert Willems and Ella Roelant
The default number of random samples is lower for GS-estimates than for S-estimates, because computations regarding the former are more demanding.
S. Van Aelst and G. Willems (2013), Fast and robust bootstrap for multivariate inference: The R package FRB. Journal of Statistical Software, 53(3), 1--32. tools:::Rd_expr_doi("10.18637/jss.v053.i03").
GSest_multireg, Sest_multireg,
MMest_multireg, Sest_twosample, MMest_twosample, FRBpcaS, ...
## Show the default settings:
str(Scontrol())
str(MMcontrol())
str(GScontrol())
Run the code above in your browser using DataLab