I read a lot of papers on the topic of "Factors Impact on Employee Satisfaction", which first proposed some possible factors, test the unidimensionality of a set of items belong to each specific factor, then run the EFA to see the proposed structure is correct and find the representative items for each factor listed on the model. This article, for example (Article The Factors Impact on Employee Satisfaction in Work at Vietn...
) . So for me, it has become the norm to have a structure in mind, Cronbach's alpha test then the EFA. But recently I came up with a lot of paper, which claimed a reverse workflow so it confused me a lot.- Article The Use of Cronbach’s Alpha When Developing and Reporting Re...
From what is written here, I understand that most of the cases, EFA is performed first then Cronbach's alpha. Depend on the number of concepts of constructs in the measure, the author can run Cronbach's alpha for the whole scale at once or factor-by-factor (see more atArticle Making Sense of Cronbach's Alpha
).So what made the difference in the 2 situations? When I should run Cronbach's Alpha first and reversely?
Thuy,
A general procedure to use Cronbach's alpha with exploratory factor analysis (EFA) follows. Once the items have been created, and data have been collected, the EFA is conducted. Once the factor structure has been established, the Cronbach alpha analyses will be used to determine the reliability of the items on the individual factors or subscales. You're generally looking for values of at least .7. If it's much lower than this you may consider removing items.
This becomes especially important when constructing an instrument, or maybe even if you're using a validated instrument in a novel sample. Many statistical packages offer an option for the alpha if an item is removed. If validating a new instrument, you can then redo the EFA without the "poor" item, and compare the results to the previous analysis. If the instrument is already validated, perhaps strategies unique to the situation may need to be explored. An example would be discussing theoretical reasons as to why the item is poor within a certain sample. Certainly, other idiosyncratic situations exist. These are good reasons why the reliability analysis is an important step after doing a principal components or factor analysis.
Hi Keston Lindsay,
Yes, I know how it works, just a little confusing in the order. As I cited above I found some paper used Cronbach's alpha before EFA. Here, I'm asking what is the difference in the 2 cases. I hope that you can explain to me about that. Thanks
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running