Maximum Likelihood Estimate of Factor Analysis Model

DESCRIPTION:
Returns an object of class "factanal" representing the maximum likelihood estimate of the model.

USAGE:
factanal.fit.mle(cmat, factors, p=ncol(cmat), start=<<see below>>,
         control=factanal.mle.control(), ...)

REQUIRED ARGUMENTS:
cmat:
a correlation matrix. Missing values are not accepted.
factors:
the number of factors in the model.

OPTIONAL ARGUMENTS:
p:
the number of variables (the number of rows and columns in cmat).
start:
a matrix with 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.
control:
a list like the result of factanal.mle.control. A list that does not meet the proper criteria will be ignored.
...:
arguments to factanal.mle.control may be given individually.

VALUE:
an object of class "factanal", see factanal.object for details.

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.


REFERENCES:
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.

SEE ALSO:
factanal , factanal.object , factanal.start.mle , factanal.mle.control, factanal.fit.principal .