There's argument efa and cfa can't be performed using same dataset. Does this consider as violation to this rule? . As measurement model us also known as cfa.
The EFA is conducted on the Pilot Study data to determine the dimensionality of items measuring the particular construct. For example how many component emerged, which items fall under which component, which items have low factor loading and need to be deleted. And finally, what is the Cronbach Alpha value for the respective component etc.
Once you completed EFA, you rearrange the items based on the result. Then you collect data from the field (field study data). Once you obtain data from the field, then you need to conduct CFA to validate the constructs.
Thus, EFA and CFA cannot be done using the same data set. One is using the Pilot Study data while the other is using Field Study data.
You can use the same data in both EFA and CFA. This depends on the theory, the purpose, and the model that you are testing.
I am attaching example of publication that used both types of factor analysis with SPSS and AMOS, hope this will help.
Regards
I think you have done EFA in SPSS and you have arrived at some factors.
In many of the statistical guidelines it is clearly given that in order to confirm the factors arrived through EFA, CFA should be done either through LISREL or AMOS.For that sake you have to treat the same factors converged in EFA for CFA .
Dear Thien,
Usually several scenarios can happen:
1- You perform a CFA and you reach the same factorial structure – you conclude by the replication of the study
2- You perform a CFA and the model doesn´t fit. Then you should to re-specify your model. You can do a EFA and use the new arrangement of items in your study .
3- You go directly for the EFA (your situation), justifying that is a new measure in your population. In this situation, ideally, you randomly split your sample into a half (or 60% 40%) and you perform with the first half of the sample the EFA and the CFA with the second half of the sample
Good luck with your research!
Hello,
I think it is not appropriate to use the same data for EFA and (subsequently) CFA for a conceptual reason. First you ask your computer to show the structure of some data (EFA, empirically-driven) and then you use the same data structure to confirm itself (CFA, which should be theory-driven). So rather than testing an apriori hypothesis, you (in a post-hoc manner) just confirm the structure you have already seen. EFA and CFA are associated with different research questions / problems. Also see "Can we do exploratory and confirmatory factor analysis in the same data set?" thread at this site (there are also some opposing standpoints).
Best regards,
M.
EDIT: I am really not sure about the splitting suggestion since the random split should create two equivalent halves of cases that "behave" much the same as the original full data. This cross-validation does not resolve the conceptual problem.
I agree with Marco. If you have a conceptual approach to the phenomenon that is trying to measure, then you should use CFA,
The EFA is conducted on the Pilot Study data to determine the dimensionality of items measuring the particular construct. For example how many component emerged, which items fall under which component, which items have low factor loading and need to be deleted. And finally, what is the Cronbach Alpha value for the respective component etc.
Once you completed EFA, you rearrange the items based on the result. Then you collect data from the field (field study data). Once you obtain data from the field, then you need to conduct CFA to validate the constructs.
Thus, EFA and CFA cannot be done using the same data set. One is using the Pilot Study data while the other is using Field Study data.
CFA on the same data set can be done - to establish convergent and discriminant validity. However, it is advisable to follow it with a CFA on a different data set to establish stability of the scale across different samples.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis