ABSTRACT
The estimation problem in factor analysis has been considered by many researchers during the last 50 years. Before the mid-sixties the predominant method for estimating factor loadings was based on some form of iterative principal components of the correlation matrix with communalities in the diagonal, which we now understand as unweighted least squares (ULS). After the mid-sixties, the maximum likelihood (ML) method due to Lawley (1940) has attracted considerable attention. For both ULS and ML several alternative algorithms exist.
