Hallo!
I have a questionnaire with 24 items. The EFA (first sample, n=490) confirmed the single factor solution of the original version (Cronbach's Alpha = .91).
The CFA (second sample, n=340, Cronbach's Alpha = .95) in AMOS has the following fit indices: CFI = .865, RMSEA = .090, TLI = .852
I know these do not meet the usual cutoff criteria.
But are they still acceptable?
Or what could I try to improve them?
Eliminating items or correlating errors does not improve the fit.
Are the above mentioned two samples homogeneous?
You may run efa on second data to see the issues.
Dear if you can slightly reduce the RMSEA to .08 or less, otherwise acceptable.
You can read my papers where I have mentioned different values to be accepted.
https://www.worldscientific.com/doi/abs/10.1142/S1363919618500573
http://www.mdpi.com/1911-8074/11/3/35/pdf
Article The role of personality in SMEs internationalization: empiri...
Good Luck
Hi,
Is the 2nd sample checked for quality in terms of structurally unplausible outliers?
Given the sample size and df, the model fits don't look like artifacts to me at first glance...
Do I understand correctly that both 1st and 2nd sample were collected in the same language version and under similar context conditions (in terms of country/region/and with no systematically differing respondent-characteristics?)
Best,
Stephan
hi!
Thank you for your responses and the information provided!
I also ran an EFA on the second sample - it showed a clear single factor solution, cronbach' alpha was .95.
I also did not find any structurally unplausible outliers.
Both samples were collected in the same language version and under similar conditions. However, the first sample included participants from the normal population while the second sample included patients.
Best,
Michaela
In my opinion the results are because of the difference in sample characteristics. I can guess the factor loadings are not as strong as it was for the first sample.
As a non-expert in your field I would suppose that systematic differences in the response behaviour between normal population and patients are a common phenomenon!?
Is there any existing research available where the same instrument was used in both a normal population and a patients sample? Would be interesting to compare...
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling