for CFA no need to go for EFA. CFA is an integrated approach and GOF indicators are enough for it. Then focus on factor loading and go for SEM. If initial model needs modification then go for adjustment maintaining the accepted values of GOF indicators.
The choice to use EFA and CFA is not subjective but rather depends on the constructs that you use in your study. In case, your model includes constructs that are not well tested in the literature (for the same country or industry) then you must apply EFA before CFA in order to check for validity and reliability of the data. However, if your constructs have been previously tested then go you can directly perform CFA.
I would not recommend performing an EFA and a CFA on the same data. If you are using a scale on a similar population that has been published upon, then you may want to go down the CFA route. If you are developing your own scale, have changed the wording of the questions, or the target population, then I would suggest going down the EFA route.
Dear all thank you for wonderful responses. Based on literature review, I have a conceptual framework consisting of 26 different latent variables, under 5 domains or themes, which are related to quality of some learning experience. For those 26 latent variables, I have 100 items. After collecting data, what would be right choice or route, EFA leading to CFA and finally SEM, or I should directly move to CFA and SEM? Considering 26 latent variables and 100 items, what problems can I face while running it on SPSS Amos? Such a number of latent variables and items would affect the model fitness? If so what precautions can be employed?
Talat Waseem, I can't get a clear grasp of what you are trying to do, but given you don't seem to have a clear/definite idea about how your conceptual framework will be evidenced in real-life data, I suspect you should begin with exploratory factor analysis (EFA) first, rather than confirmatory factor analysis.
However, if you have 100 items you might need about 1,000 participants in order to conduct EFA. Are you likely to be able to access such a large number of participants? If not, I think you should explore ways in which you could reduce the number of items you want to analyse.
If you have a model that assigns each of your 100 items to one of your 26 latent variables, and each of those latent variables to one of the 5 higher-order domains, then that is definitely a specification for CFA, and there is no need for EFA.
I have to say, however, that this is a very complex model and thus the chances that it will actually fit the data are rather small. So, EFA might be a useful tool for deciding how to simplify the model.
Also, since you have nearly 150 loadings and correlations, that implies you need a sample of 750 or larger to get stable estimates.
In my experience, if you've theoretical backing for a model including 26 latent factors, and corresponding items are adopted from already published scales/instruments, you should go for CFA. However, as mentioned by Robert and David, you will need to big sample size to run CFA for 26 variables altogether.
Further, since you mentioned SEM and the AMOS, I believe you are talking of SmartPLS SEM in former case. This also, kind of, points that you don't have a big sample size either ( pure assumption based on your query). So, if sample size is not that big, I would recommend dividing the model into smaller model (of course supported by theory). Handling 26 latent variables in AMOS will be challenging, not because AMOS cannot handle it but because of poor GUI of AMOS. I will definitely go SmartPLS SEM with as many variables as yours.