
Message/Author 


Hello there. I am trying to get some clarity on the KaiserGuttman (eigenvalue) rule for EFA. I have seen in the literature examples where the authors adjusted the cutoff downwards due to communality estimates inserted on the diagonal of the correlation matrix (the analysis was a principal factor analysis). I have also read an article that suggested it was not even appropriate to use this criterion at all in the context of common factor analysis. The Mplus Discussion boards seem to suggest that the standard cutoff of 1 is appropriate... 1. For an EFA of categorical data using the ULS estimator, is the eigenvalue1 rule still appropriate? 2. If so, is it common  or does it make sense  for this criterion to suggest retaining many more factors than would be suggested by, say, requiring RMSR < 0.05 or a scree plot? Many thanks in advance! 


For background on these matters, the following article is helpful: Fabrigar, L.R., Wegener, D.T., MacCallum, R.C. & Strahan, E.J. (1999). Evaluating the use of exploratory factor analysis in psychological research. Psychological Methods , 4, 272299. 1. The eigenvalues that Mplus prints refer to the sample correlation matrix, so with 1's on the diagonal. The eigenvalue1 rule is intended for such matrices. 2. I don't know how common it is but I prefer to use a scree plot coupled with interpretability. See also other methods in the above article. 

Back to top 

