Hi all, Need help for scenario .
I have total sample size of 800 with 6 constructs and 3 questions in each questions . After doing factor analysis i am getting Cronbach's alpha for 2 constructs as 0.98. Kindly suggest how to defend such high value of Cronbach's alpha?
Regards,
Hello Uday,
With three questions per latent variable (I presume that's what you meant in your query), it would be a bit unusual to obtain such high values for internal consistency score reliability.
If you're first factoring the question set, and then evaluating the estimated internal consistency score reliability, you may be getting an overestimate (if individual questions/variables have been selected or retained based on their loadings on the respective construct).
There's a classic paper by Cureton which addresses this tendency (albeit in the reverse direction): Cureton, E. E. (1950). Reliability, validity, and baloney. Educational and Psychological Measurement, 10, 94-96. doi:10.1177/001316445001000107 (link: http://www.stat.columbia.edu/~gelman/stuff_for_blog/Cureton_Baloney.pdf).
Another possible explanation is that the questions for a given construct overlap very strongly to the point of redundancy (e.g., "I like chocolate ice cream;" "I am fond of chocolate ice cream").
A final possibility is that the participants weren't particularly forthright in their responses, and chose similar options for all the questions/variables in a group, so as to get finished more quickly.
Good luck with your work.
α = 0.6-0.7 indicates an acceptable level of reliability. The value of 0.8 or greater values are very good level. On the other hand, values higher than 0.95 are not necessarily good, since they might be an indication of redundancy. Please also consult SPSS Manual.
Hello Uday
you state that you have two (latent) variables with three items each, however, you only gave us one figure for Cronbach's Alpha (.98). Is it .98 for both variables? If you have not done so already, compute the Alpha value for both variables seperately or even better: drop Alpha for Omega (available in SPSS 27).
Check out these publications:
Article Use Omega Rather than Cronbach’s Alpha for Estimating Reliab...
Article Thanks Coefficient Alpha, We’ll Take it From Here
If that is not an option, check the inter-item correlations between the three items for each variable. I would image they are quite high which indicates redundancy, as David has pointed out. Also, I am fond of chocolate ice cream, David Morse .
Best
Marcel
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression