John Venz posted on Monday, August 10, 2015 - 6:57 am
Hello, I have a few questions regarding factor scores. (1) When I use the FSDETERMINACY option in Mplus, I get a Sample Covariance and Correlations Matrix of the estimated Factor Scores. These matrices have TWO lines and TWO rows for each latent factor and hence 4 cells for every crossing betwenn two latent factors. I figured out which one is the empirical covariance but what do the other numbers mean? (I'm using Mplus 6.11.) (2) Are the calculated Factor determinacies the same what is often called 'rho' in this context, i.e. the multiple correlation between each factor and the original variables, e.g. in the popular paper by Grice(2001)? If yes, are the sqares of these values a measure for the amount of the 'determinate part' of the factor scores, as suggested by Grice(2001)? (3) Grice argues that a determinacy value lower .707 is 'very bad' since it means that sets of factor scores for the same factor that are even orthogonal or negatively correlated are possible. Would you go along with this? (3)a When Factor determinaicies are around .78, say, can I still use Factor Scores as independent variables in subsequent Regression Analysis or are determinacies 'too low'?
Any help is kindly appreciated.
Grice, J. W. (2001). Computing and evaluating factor scores. Psychological Methods, 6(4), 430–450.