I have collected data on 6,7 variables by using already established questions. Now in analysis section, I have to report validity of the measurement instruments. Either (i) I should use principal Component Analysis (PCA) (a form of EFA) for each individual variables as questions are already established and my agenda is not to identify dimensions or use (ii) CFA for validity reporting.
In research papers, I have seen both the techniques used without any explaination.
(I) Are these approaches the alternatives to each other?
(2) What are the merits and/ or demerits of each?
(3) And acceptable/ standard/ banchmark values of factor loading?
If your measurement instruments have already been proven valid in your local setting, you don't have to re-run factor analysis. Just cite previous validation studies. EFA is more of trying to determine the underlying constructs of the measurement tool, whereas CFA sees how well it fits the theorised constructs. However, you would need to determine the internal consistency which would change according to different settings. As for factor loading, acceptable cut-offs are 0.4 and more. However, some choose a stricter level of 0.5 and above.
For proven construct, only CFA is necessary.
For the standards, measurement model (CFA) and structural model use exactly the same criteria (see Hu and Bentler (1998) ; Martinez-Lopez et al. (2012)).
good luck!
The answer is CFA if you are using pre-established scales. However, I could see that you have 6 or 7 variables. The use of exploratory factor analysis is not befitting in this research as the number of variables is very less. Also, the type of variable must be continuous and scale based if you wish to run either method. The objectives and questions used would give more clarity to help you.
First, do EFA for factor analysis then do CFA for confirmation of the model and factors.
If you are using an existing instrument, you have only to check whether it is valid at your setting. I will try to establish the validity of each individual variable (construct) through CFA.
EFA is for determining the dimensionality of items only. CFA is for validating the constructs.
If you are talking about validating the construct, you should employ CFA because through CFA procedure you could validate the construct for Construct Validity, Convergent Validity, and Discriminant Validity. Also for Composite Reliability.
I echo Chai. At least in my discipline, for previously validated and used scales you will only need to cite prior studies, check for internal consistency and include Chronbach's alpha value.
You can directly go for CFA as the scale is already established one. Still, to be sure, you can run EFA if the same scale is being contextualize in your setting for the first time.
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology
- Research Papers