In determining internal consistency reliability for measurement models, composite reliability values of more than 0.9 is deemed not desirable because this indicate "they are measuring the same phenomenon and are therefore unlikely to be a valid measure of the construct" (Hair et al., 2017)
Hair, J. F., Hult, G. M., Ringle, C. M., & Sarstedt, M. (2017). A primer on partial least squares structural equation modeling (PLS-SEM) (2nd ed.). Thousand Oaks, CA: Sage Publications.
However, discussions in forum have noted that values above 0.9 is acceptable, as it affirms unidimensionality of the construct, as long as items in the scale are not redundant http://forum.smartpls.com/viewtopic.php?f=5&t=3805 .
May I know is there any literature which explicitly addresses this issue ?
Here's a reference for high Alpha (the older cousin of CR), suggesting that above 0.9 may point to redundancy: Article Making Sense of Cronbach's Alpha
Now, this is quite an established fact but the forum discussion you shared gives a good insight into the highlighted "may". Dr Becker is saying, in my understanding, that items should measure different aspects. If they are, specially based on theory, and you still achieve a high number, it shouldn't be a problem. On the contrary, if the items are measuring same thing over and over again, you will achieve a high figure but in a 'redundant' fashion.
Catch is the subjective review of the items which are achieving such a high figure. No standard answer applies to all situations.
Hopefully, it helps.
Try to drop the items that are, theoretically, 'very' similar and lower alpha's value (scale if item deleted) before re-running the reliability test again.
Hi Yim,
You may refer to the latest discussion concerning the same topic at
https://forum.smartpls.com/viewtopic.php?f=5&t=3805#top
Hi Saeed,
I've noticed this development, thank you for updating this space
The logic behind reliability analysis is to check the inter correlation value between the items of a construct.
Therefore a value of .9 or more acceptable. But if the value if near to 1 the test shows that both the items has the same inter correlation value. In such cases. delete the one fo the item.
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publications