fwdmv.init {Rfwdmv} | R Documentation |
This function computes a multivariate forward search for ungrouped data. Several diagnostic statistics are monitored during the search: see fwdmv.object
. Note that this function is called by fwdmv
when no tentative groups are specified. It is recommended that fwdmv
be used for all multivariate forward searches.
fwdmv.init(X, bsb = ellipse.subset, scaled = TRUE)
X |
a matrix or data frame containing the multivariate data set. |
bsb |
a function of two variables: the multivariate data in matrix form X
and the number of units in the initial subset size .
If bsb= bb.subset the initial susbet is found using robust bivariate boxplots. The default is to
use robustly centered ellipses ellipse.subset .
Alternatively, the initial subset my be specified directly by providing an integer vector
containing the indices of the units to be in the initial subset. |
scaled |
a logical value. If TRUE then scaled Mahalanobis distances are used during the forward search. |
This function computes the Forward Search described in chapter 3 of ARC. The initial subset can be
specified directly in the argument bsb
or computed from the data.
By default bsb
is a function for computing the initial subset using robustly centered ellipses.
Given a subset of m
units the next subset is the m+1
units with smallest Mahalanobis distances calculated using the center and covariance matrix
estimates of the units currently in the subset.
This process is repeated until the subset contains all of the units and several diagnostic
statistics are computed for each subset.
aa fwdmv
object.
Kjell Konis
Atkinson, A. C., Riani, M. and Cerioli, A. (2004) Exploring Multivariate Data with the Forward Search. Springer-Verlag New York.
data(fondi.dat) fondi.init <- fwdmv.init(fondi.dat) data(fondi.dat) #### find the intial subset using robust bivariate ellipses #### start with an initial subset size of 17 units fondi.init <- fwdmv.init(fondi.dat,bsb=ellipse.subset(fondi.dat,17)) data(fondi.dat) #### find the intial subset using robust bivariate boxplots and #### start with an initial subset size of 17 units and fondi.init <- fwdmv.init(fondi.dat,bsb=bb.subset(fondi.dat,17))