Hi ? I ran CFA in AMOS and fit indices were poor . I ran EFA in SPSS (etracted six factors) and delete some Items that were cross-loading , then run CFA again and now fit indices are within acceptable range ( only TLI failed =087). SRMR =055, CFI=093, RMSEA=065 and CMIN/DF
Dear Bruce,
first things first:
1. How did you design your EFA (sample size, PAF or PCA, rotation method, extraction etc.)?
2. Did you run your EFA and CFA with the same sample? If so, choose either EFA or CFA. You cannot explore your model with the same sample you use to validate it. That would be a self-fulfilling prophecy.
-------
3. to your fist question: No. You get a second-order model/factor when you define, say, three CFA-models, by another factor. Please see the sketch attached. Also, this is just a random model I found on RG that shows a second-order factor: Article Usefulness of WISC-IV in Determining Intellectual Giftedness
4. None of your fit-indices look really terrific. CFI and TLI usually go together, so I would not drop them. However, without knowing your data it is nigh on impossible to say how to improve the fit-statistics.
Hope that helps at least a bit.
Kind regards from Germany
Marcel
Dear Bruce:
Regarding your first question, second-order construct is when you have a construct pointing to other construct. Second-order constructs are used to overcome multicollinearity issues and to make the model more parsimonious.
Bests
You should NOT start with a CFA. The proper procedure is to run EFA first on one set of data. Then you run CFA on a second set of data to validate the factor structure of the EFA. If the outcome of the CFA is not ideal, you still can modify the model to improve the fit indices. You may need to run CFA for a few times depending on the number of the modification.
CFI, SRMR, and RMSEA should be reported. TLI is optional since not everyone reports this. Chi Square/df is not reliable since it is affected by the sample size.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running