Dear All
I have conceptualized a parallel mediation model which has not been tested before. I used EFA to reduce the number of items from 31 to 22 and then performed CFA with SEM. The final number of items were 14. However, since EFA is performed when factor structure is unknown, is it justified to say that EFA was performed to reduce the number of items and check the factor loadings, while deleting those with multiple crossloadings and low values?
@Oshin Indeed EFA is used to reduce the dimensions of the data. As you stated your case, after EFA, number of observed items reduced to 31 to 22 which seems a bit discouraging if your used published and validated scale. Further, in CFA, you again lost eight items due to low loadings which indicates towards some problems in your survey instrument or sampling design during data collection. Usually, these many variables in EFA and CFA are supposed to be drop when a new scale is being developed and validated. While establishing the reliability and validity of the data using already validated and published scales, only a few items are dropped not these many.
Dear Oshin Gulsia
Exploratory factor analysis (EFA) is generally used to discover the factor structure of a measure and to examine its internal reliability. EFA is often recommended when researchers have no hypotheses about the nature of the underlying factor structure of their measure.
Best Regards
Dr. Fatemeh Khozaei
Thank You Bilal Mustafa Khan Fatemeh Khozaei Imran Anwar for your valuable suggestions. Regarding scale development, I would further like to ask if I am using four different scales for four different latent variables in my model with little modifications as after pretesting, is it justified to use an EFA then?
@Oshin As you mentioned that you've made modifications to adapt existing scales as per your research settings, in this case, you should confirm the structure of these scales using EFA then go for CFA.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling