is convergent validity, i.e. average variance extracted and composite reliability, is always necessary in structural equation modeling?
As we know that CFA is part of SEM, to validate the scale validity, can we use international consistency alpha values, in addition to AVE and CR?
anyone knows some articles saying that AVE and CR must be done or some articles saying that AVE and CR are not always necessary?
As far as I know, CR and AVE are always computed to guarantee the validity of the structural model.
If you are conducting CFA for testing the validity of the scale, the AVE can be used to assess the convergent validity of the scale. On the other hand, the internal consistency reliability or Cronbach's alpha is an indicator of the consistency of the items in the scale. In a reflective scale, the internal consistency reliability of the scale is assessed through the composite score of Cronbach's alpha. However, if the scale is a formative scale, Cronbach's alpha of each dimensions should be assessed separately.
Hi Yuting;
you can watch this video
https://www.youtube.com/watch?v=yk6DVC7Wg7g
and you can use this excel sheet
http://www.kolobkreations.com/Stats%20Tools%20Package.xlsm
to calculate discriminant and convergent validity. This excel sheet has formulas about this type of validity.
- Hair et al (2010) page 618-620 may help to understand this concept.
Hair, J., Black, W., Babin, B., and Anderson, R. (2010). Multivariate data analysis (7th ed.): Prentice-Hall, Inc. Upper Saddle River, NJ, USA.
Thank you for all the answers. After checking some papers, I found there is no agreement about what measures to use for the scale reliability in CFA/SEM. Some one used Cronbach's alpha, some one used AVE and CR. I do have a factor which has a high Cronbach's alpha value, but AVE is 0.34 and CR is 0.66, which do not meet general requirements of AVE (>=0.5)and CR (>=0.7). Does this mean this factor is not valid? I don't think so.
Anyone can suggest some books about scale reliability measures?
Hi Chen
In your case, AVE and CR both are low, hence I would suggest recheck the factor's items and their respective loading. Discard the items with poor loading, it might improve AVE and CR as well. And if your study is an exploratory kind of (scale is not standard and developed by the author), an AVE below .50 could be considered if CR is above .70 ( Forrell & Lacker, 1981). This might give support to retain the factor.
Follow:
https://www.researchgate.net/post/Is_it_possible_to_continue_with_a_low_average_variance_extracted_AVE_value
In my case, I have composite reliability (CR) 0.77 and Analysis of Variance Extracted is 0.45. does it mean the scale is valid and reliable.
Ijeoma in your case you are just lucky. CR ,=.77 and AVE .45 is justifiable and can be retained. There are a lot of readings ascertaining it
@ Lutete Christian Ayikwa can you please suggest me some research articles which provide support for AVE .45 is acceptable
Zahid Hameed you need to visit following book where alternative methods are give alternative measure for validity:
Malhotra N. K., Dash S. (2011). Marketing Research an Applied Orientation. London: Pearson Publishing.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts