I have run the CFA with 7 latent variable ( Customer orientation, competitor orientation, inter-functional coordination, resource, low-cost, differentiation, and performance. However, there is a multicollinearity between customer orientation and resource. How Can I handle with this? or should I delete one of latent variable?
Thanks
You may want to consider EFA to identify the number of latent factors.
If the model with 7 latent factors you mentioned is theoretically supported, you can run another CFA on a model with just 6 latent variables (i.e., combining customer orientation and resource), compare the two models, and see which model fits the data better.
Hi Ho Le Phi Khanh,
Agree with Mr. Chee-Seng Tan. However, You may first confirm normality test with both univariate normality using Skewness & Kurtosis value and Multivariate Normality using Mahalanobis distance. After confirming both Univariate and Multivariate normality of all of your items, then run EFA and check for dimensions. If EFA still provides six or seven dimensions, then RUN again CFA and check the results. (Factor loading value, AVE, and Correlation statistics )
If still your construct has multicollinearity problems, then you might need to discard the latent construct or if the construct is derived from theory and literature argues very important construct then you can continue with this one and mention it in the limitation section.
With Thanks
Moniruzzaman
You can create a secondary factor model with customer orientation and resource as indicators. Whatever you do, make sure that it is theoretically sound!!!!!!! Here's how: https://www.youtube.com/watch?v=HBQPqj63Y7s
Even if two measures have high correlation but if the discriminant validity is proved then it is not the issue. However I do agree with Mr. Chee-Seng Tan that you can run EFA first to identify the factors and then apply CFA to confirm the same.
Regards
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude