factanal
representing the maximum likelihood
estimate of the model.
factanal.fit.mle(cmat, factors, p=ncol(cmat), start=<<see below>>, control=factanal.mle.control(), ...)
cmat
).
p
rows and an arbitrary number of columns, each column of
which is a starting value for the uniquenesses.
The default is the result of
factanal.start.mle
.
factanal.mle.control
.
A list that does not meet the proper criteria will be ignored.
factanal.mle.control
may be given individually.
"factanal"
, see
factanal.object
for details.
The algorithm is a modified version of that described by Joreskog (1977),
which is essentially a Newton-Raphson procedure with some tricks specific
to this estimation problem.
The main modification from the Joreskog algorithm
is that the solution is constrained to remain
strictly within the allowable region.
The constraint allows estimation to proceed when Heywood cases occur.
The algorithm tests each of the starting values given in
start
and
uses the one with the largest likelihood.
Joreskog, K. G. (1977). Factor analysis by least-squares and maximum-likelihood methods. In Statistical Methods for Digital Computers. Enslein, K., Ralston, A. and Wilf, H. S. (editors). Wiley, New York.