procrustes
,
promax, the
oblimin family, and the
orthomax
family.
factanal
,
princomp.
rotate(x, ...)
rotate.default(x, rotation=NULL, orthogonal=T, parameters=NULL,
normalize=T)
rotate, and a matrix (generally with
fewer columns than rows) for
rotate.default.
"varimax",
"quartimax",
"equamax",
"parsimax",
"orthomax",
"covarimin",
"biquartimin",
"quartimin",
"oblimin",
"crawford.ferguson",
"procrustes",
"promax",
"none".
If
rotation is
NULL, then the remaining arguments define the rotation.
TRUE, the rotation is constrained to be orthogonal.
"oblimin",
"orthomax"
and
"crawford.ferguson") will use them. This is required for Procrustes
rotations and must be a matrix of the same dimension as
x.
If
parameters is
NULL, then the default for orthogonal rotations is
parameters=1, and for oblique rotations the default is
parameters=c(1/p, 1, -1/p, -1) where
p is the number of rows of
x.
TRUE, then Kaiser normalization is performed.
rotate that is used.
rotate.default returns a list which depends on the type of rotation
(see
obliquemin,
orthomax and
procrustes). The list that is returned
will always contain the following components:
rmat up to numerical precision.
The
parameters argument provides a means of selecting a specific rotation.
For orthogonal rotations (
rotation=orthomax, or
rotation=NULL and
orthogonal=T
),
parameters must be of length 1.
For the oblimin family,
parameters should be of length 1 or length 4.
Crawford-Ferguson rotations allows
parameters to be of length 2 or 4.
Harman, H. H. (1976).
Modern Factor Analysis,
3rd Edition.
University of Chicago Press, Chicago.
Mardia, K. V., Kent, J. T. and Bibby, J. M. (1979).
Multivariate Analysis.
Academic Press, London.
amat <- matrix(c(2,-1,-2,3,2,1,-2,-4),4) rotate(amat) # varimax rotation rotate(amat, "quartimin") rotate(amat, ortho=T, param=10)