Information Technology Reference
In-Depth Information
4.7 Functions for constructing a CVA biplot
The main function for constructing CVA biplots is the function
CVAbipl
. Its usage and
arguments are similar to those of the function
PCAbipl
as discussed in Section 3.7, with
notable exceptions given in Section 4.7.1. There are also several functions closely related
to
CVAbipl
for constructing specific types of CVA biplots or adding enhancements to
an existing biplot. These functions are briefly introduced in Sections 4.7.2
−
4.7.7.
4.7.1 Function
CVAbipl
This is the basic function for constructing CVA biplots. Provision is made for one-, two-
and three-dimensional CVA biplots. Its call and arguments are almost identical to those
of the function
PCAbipl
as discussed in Chapter 3, with the following exceptions:
Arguments
A required argument for
CVAbipl
. Indicator matrix
defining
K
groups of samples in the data.
G
Not available in
CVAbipl
since CVA is scale-invariant.
If for some reason a CVA of scaled data is required it
can be performed by assigning the scaled data to
argument
X
.
scaled.mat
The default is
"weighted"
, specifying a weighted CVA
to be performed. Other possible values are
"unweightedI"
or
"unweightedCent"
for
specifying the different forms of unweighted CVA
discussed in Section 4.2.
weightedCVA
Confidence coefficient, typically 0.95 or 0.99, for
constructing confidence circles about each group
conf.alpha
mean. The factor
√
n
i
, as described at the end of
Section 4.2, is included in the calculations.
Value
This is a list with the following components:
The between sums of squares and products
matrix.
SSP.B
The within sums of squares and products
matrix.
SSP.W
Matrix
L
that transforms the original space into
the canonical space.
Lmat
Matrix
M
=
LV
.
Mmat
The eigenvalues i.e. diagonal elements of
.
lambda
The overall quality of the display in terms of
the canonical space.
CVA.quality.canvar
