4
$\begingroup$

I understand that it's instead correct to cross-validate using new data. Why is it so? It is just that a model will tend to fit the data set that was used to created it better than another randomly sampled set of data?

Could it ever be justified to use the same data for EFA and CFA?

$\endgroup$
1
  • $\begingroup$ Muthen frequently recommends a more flexible model that resembles EFA where are manifest observations are allowed to load on all latent variables to test the fit of the CFA model. See here for one example. $\endgroup$
    – Andy W
    Jun 3, 2014 at 14:52

1 Answer 1

8
$\begingroup$

It is generally a bad idea to do an EFA and a CFA on the same data for the exact reason you mention: A factor structure derived from an EFA will almost always fit very well in a CFA using the same data. EFA and CFA are closely related, so it is no surprise that this is the case.

It is common to split data in half and to do EFA on one half and CFA on the other half.

CFA could be justified on the same data as was performed an EFA if you are interested in demonstrating fit indices for comparative purposes, to examine modification indices as a mean to further elaborate on the model, and to examine covariances and such between factors. In these cases, you wouldn't do CFA to demonstrate that your model fits your data or to simply support your theory/model. This would be a very weak test of model fit because its the same data as from the EFA.

$\endgroup$
1
  • $\begingroup$ Did the beginning of the last paragraph mean "CFA [on the same data as EFA] could be justified, [though], if ..." (which I think is reasonable) $\endgroup$
    – ttnphns
    Jun 3, 2014 at 15:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.