PLS/SmartPLS and Unidimensionality

[ighulammustafa]

What if the unidimensionality criteria is not satisfied? How should one handle such a situation? From the literature,

"Unidimensionality is usually satisfied by retaining the items whose loadings are above 0.7". (Reference Needed)

"To measure the assumption of unidimensionality, Cronbach’s alpha is the most common method. The value of the Cronbach’s alpha should be greater than .7 to measure the assumption of unidimensionality." (Reference Needed)

"In factor analysis, to measure the assumption of unidimensionality, the cutoff value of the factor loading should be higher than .3, or eigenvalue should be greater than 1." (Reference Needed)

I have a construct with 5 measurement items. The loading values are 0.5282, 0.9063, 0.3489, 0.4478, and 0.477. The cronbach's alpha is 0.6346, and the composite reliability is 0.6865. If I start to exclude the items from the construct (based on the loading value < 0.7), I'm left with just one item -- the loading values don't go above 0.7 for any other item (except the one with 0.9063). How should I handle this case?

EDIT:

I have another construct with 3 measurement items with loading values of 0.2443, 0.4386, and 0.9602. The cronbach's alpha is 0.5073, and the composite reliability is 0.5965.

The model consists for 6 constructs. 2 out of 6 have unidimensionality issue.

[Hengkov]

Hi,
For development construct, outer loadings 0.4-0.5 acceptable.
But for testing must 0.6-0.7.
Check data outliers for solve this problem.
Your respondent give bad answer.

Regards,