I conducted an explanatory factor analysis (I used different rotations and used principal component). The explanatory factor analysis wasn't able to identify two of my model's constructs as separate constructs, meaning that all of the two constructs' items load on one factor only. However, that seems pretty obvious to me, since both are highly correlated, which is a good thing, since one of the two factors is an antecedent of the other factor. In other words, I want them to be highly correlated. How shall I further proceed?
Dear Maximilian,
Perhaps an alternative is to try a factor analysis technique instead of principal component analysis, such as principle axis factoring. Factor analysis techniques focus on shared variance rather than total variance. The issue is then the relative strength and independence of the second factor.
Please see my study guide.
@ Leonhard: thanks for your answer. I use PLS-SEM, which is why I cannot calculate overall model fit. However, you recommend calculating the model twice, incentive with only one construct, consisting of the indicators for both constructs and once with two constructs?
@ Peter: thanks for your answer, too. I thought that methods such as maximum likelihood or principal axis require normally distributed data? Can I combine the method with varimax or do I need to apply other rotations?
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS