Bo Fu posted on Monday, November 08, 2010 - 12:00 am
Does mplus provide contribution proportion (either separate or accumulative proportion)? The proportion is calculated based on the eigenvalue. The eigenvalues could be shown in the output, but if I do an EFA for large number variables, i.e. 200 variables, to compute by hand is hard because it's not convenient to extract those eigenvalues from output.
The main reason is that variance proportion computed via eigenvalues is relevant for principal component analysis - which has as a goal to explain variance - rather than factor analysis - which has as a goal to explain correlation. Also, with correlated factors such a decomposition of variances contributed by the factors is not possible.
Some packages report variance contributed by factors, but that is typically in conjunction with Varimax rotation (uncorrelated factors), and my opinion is that both Varimax and such reporting is suboptimal/outdated. Others may disagree.