Weighted, Multiply Imputed Data PreviousNext
Mplus Discussion > Exploratory Factor Analysis >
 Amy Schumacher posted on Wednesday, March 18, 2015 - 9:48 am

I am working on an EFA, and eventually an SEM, with weighted, multiply imputed data with ordinal variables. I have a total of 26 ordinal variables that I expect to yield up to 6 factors. Right now I'm working with a subset of 6 of those variables that should load on 2 factors to figure out the code.

I have run the EFA both with just imputation 1 with the weighting variable using ANALYSIS: TYPE = EFA 1 2;.

I have also run it with all 5 imputed data sets and the weighting variable using MODEL: f1 - f2 BY <variable names> (*1);.

My current plan is to use the standard EFA with 1 imputation to examine the eigenvalues, make a scree plot, and get a feel for how the factors are loading, and then confirm my results with the one with 5 imputations by running all the the possible contenders and comparing model fit.

Here are my questions:

The imputed analysis yields a monte-carlo type set of model fit statistics. How should I interpret these?

The imputed analysis gives me a correlation of 0.91 between F1 and F2, while the 1 imputation EFA gives me a correlation of 0.319. Why are they different?

I requested residual output for the imputed analysis, and there is a note that says NOTE: These are average results over 10 data sets. However, I only have 5 imputed data sets. What is happening there?

Thank you!
 Tihomir Asparouhov posted on Wednesday, March 18, 2015 - 1:27 pm
Please send the input output and data to support@statmodel.com
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message