Hello and thank you for the opportunity to post questions. I am working on an exploratory factor analysis with categorical data (3 response options - yes/not sure/no) from a sample of nearly 3000 children. I have a couple of basic questions:
1) From perusing this site, I understand that MPLUS gives significance levels for the factor loadings, but I am not sure exactly where those are presented & how to use them - please advise a beginning user!
2) I'd like to construct a scale from the factor loadings - can I do that in MPlus and then export to Stata? Or can I use the factor loadings as simple weights and construct my own scale directly in Stata?
Thank you very much Dr. Muthen, for the rapid response. Could I ask you or any of the readers to point me to some further details in the user guide or elsewhere? Do I use the factor loading/s.e. as a z-score, and determine p-values accordingly (e.g. z-scores greater than |1.96| are statistically significant?
Second, is it possible to construct the scale in MPlus using the EFA analysis?
See the OUTPUT command chapter where this is deacribed.
You can use ESEM to save factor scores. ESEM is the same as EFA if there are no other parameters in the MODEL. See the MODEL command and also examples at the end of Chapter 5.
Sarah Wu posted on Tuesday, March 10, 2015 - 1:37 am
Hi Dr. Muthen,
To add to this discussion, how do using significance values to the parameter estimates compare against using cutoff values (i.e., a loading of at least .60)? I haven't found many papers on EFA that discuss using p values for factor loadings. Thanks!
Cudeck, R., & O’Dell, L. L. (1994). Applications of standard error estimates in unrestricted factor analysis: Significance tests for factor loadings and correlations. Psychological Bulletin, 115, 475–487.
See also our ESEM article
Asparouhov, T. & Muthén, B. (2009). Exploratory structural equation modeling. Structural Equation Modeling, 16, 397-438. Contact the second author. Paper can be downloaded from here.