Yoon Young posted on Monday, October 08, 2018 - 6:44 am
Dear Dr. Muthen,
Below is my EFA results. None of the p-values are non-significant. The parallel analysis suggested 4 factors, the eigenvalues suggested 6 factors, and the model fit comparison seems to extent this to 7 factors. Would you guide me which criteria I should consider as a priority in this case? Thanks a lot!
SUMMARY OF MODEL FIT INFORMATION
Number of Degrees of Model Parameters Chi-Square Freedom P-Value
These conflicting messages often happen in real data because the data may not represent a perfect factor model with a certain number of factors. For instance, you may have m factors but also several residual correlations. This then leads to chi-square saying you should have more factors than m - which is not the true model. This often shows up as some factors having only 2 significant loadings. Look at modification indices to see if residual correlation is present. In general, parallel analysis may be more robust to this, but I am not sure this has been thoroughly studied. And of course you need to see how any solution relates to theory for the topic.
Yoon Young posted on Tuesday, October 09, 2018 - 11:23 am
Thank you so much! It is my first time examining the modification indices so would you please guide me how/what to look at?
Here is the part of the output.
MODIFICATION INDICES FOR ANALYSIS WITH 4 FACTOR(S)