James Yonker posted on Wednesday, December 31, 2008 - 3:54 pm
It is my understanding that a measure of factor determinacy is only available for factor models with all continuous indicators. Is there any analogous measure for a factor model that includes ordered categorical indicators? If yes, what is this measure, can it be computed from Mplus output (automatically or via programming), and what is the interpretation? If not, could you please explain why such a measure is not available? (I looked at Technical Appendix 11, but it isn't clear to me exactly why such a measure isnít computed. Is it because of the assumption of no residual error?)
I am essentially trying to gauge the quality of computed factor scores for use in subsequent analyses by other researchers. Any advice or suggestions would be helpful and welcome.
Factor determinacy is printed for categorical outcomes by default when using EFA. The interpretation is the same as with continuous outcomes, reflecting the correlation between the estimated and true factor scores. But you are right that using information functions is a more common approach to gauging the quality of estimated factor scores with categorical outcomes. Those functions give the inverted SE^2 for the estimated factor score at different true factor score values. This is described in our IRT tech appendix,