Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It only takes a minute to sign up.
Sign up to join this community
How meaningful or informative is it to follow an Exploratory Factor Analysis with Confirmatory Factor Analysis that uses exactly the same indicators and factors?
My goal is to use various CFA methods (such as ML, Satorra-Bentler, Yuan-Bentler, Bollen-Stine, WLS and DWL) to demonstrate in detail how well the model specified by the EFA fits the data.
Do you think this is a good idea? Or is it unnecessary?
Please advise.
Thanks.
An EFA has all indicators load on all factors and will always fit the data better than a CFA which constrains some loadings to 0. CFA is thus a stricter test and if it fits well enough, will provide stronger evidence than EFA.
It is common to use EFA to select an appropriate number of factors and to identify the pattern of loadings, followed by CFA with the same number of factors and with all the EFA loadings under some threshold (e.g., 0.3 or 0.4) constrained to 0. This CFA is sometimes done on the same data as the EFA and is sometimes done on independent data. It is more rigorous to use independent data and this helps to attenuate "overfitting" but there is still some value in using the same data as this will show how robust the model is to the aforementioned constraints.
