I need a help to understand when actually can do exploratory factor analysis? And is it possible to do it for one construct that have dimensions, ie. HRM practices includes ( training & development , performance appraisal, compensation ). Is it possible to test exploratory factor analysis on those dimensions one by one instead of doing all HRM practices items together, if yes why? And what is the justification should be
Shaker,
In my practice, I seldom start with EFA. If I suspect my instrument of being multidimensional, I'd start with the CFA based on a priori structure (I have trouble imagining not having one). A strict simple-structure CFA, though, often has fit difficulties. First, don't be afraid of including cross-loadings a priori. Second, a Bayesian CFA (implemented at least in Mplus, and most likely in R) can help you assess how far off from a simple structure you are.
If, after a Bayesian CFA and logical (and preferably anticipated a priori) changes, the model still does not appear consistent with the data, *then* I might throw my hands up and go with EFA to find out where I'm badly mistaken.
This is not the only approach, of course. Mulaik's "four-step" procedure includes an EFA as an early step, but to determine the appropriate *number* of dimensions, not their content.
Hope that helps,
Pat
Dear, Patrick and Jan Böhnke,
Thank you for the quick and helpful answer, but to be frankly with you , i am Phd Student i have done my viva successfully with major correction, which was because 2 things :
first because i forgot to make a pilot study before collecting my data which mean i have not test the validity and reliability for the study measurements before go for collecting and analyzing data . and the reason in my thought was because those measurements was tested by previous studies and it was valid and reliable but the examiner mentioned that even though i must do pilot study since i am conducting the study in other country or environment so i dont know whether there is a way to solve this problem or not without recollecting my data again and start from beginning which i have no time for that.
second because factor analysis. i have used exploratory factor analysis and during this stage i have put all items of HRM practices including ( training & development , performance appraisal, compensation )" together. and what i found later that some items with high loading come under other dimension ( i.e some of items from training and developments come under compensation ) and many items has been deleted. so i realized that CFA is the right way but the problem now i dont know how to use MAOS and dont know how to test through CFA, it will take time to learn so i decided if i can skip the problem by using EFA but dimension by dimension ( BUT STILL NEED TO JUSTIFY WHY WHICH IS I DONT KNOW) i will appreciate so much if anyone can suggest any solution or support my thought.
thank you
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology