Hi!

I have run a structural model and am interested in deriving relative importance for each of the attributes that drive interest in my ultimate dependent variable. Is this possible to do?

Below is the model summary -

ATT1, ATT2, ATT3, ATT4 formed a latent variables L1

ATT5, ATT6, ATT7, ATT8 formed a latent variables L2

L1 and L2 are correlated and are predictors of DEP

How can I find out the contribution of ATT1 to ATT8 in predicting DEP? Please let me know. Thanks in advance for your help.

Regards,

LearnSEM8

My view is that you can't, but for a relatively interesting reason.

You state that "ATT1, ATT2, ATT3, ATT4 formed a latent variables L1", but that's not strictly correct. Variables ATT1-ATT4 are caused by or predicted by latent variable L1. You won't find the contribution of an ATTx variable predicting DEP because both the ATTx and DEP variables are dependent, and share common causes L1 and L2. The latent variables both predict DEP and explain why the ATTx variables are correlated.

However, the factor loading pattern can tell you a lot about both the relation of each item to its relatively factor and the model implied covariance between each item and DEP. Higher factor loadings denote higher relations between an items and the set of other items & DEP. Similarly, the model implied covariance between any item and DEP is going to be loading*factor variance*DEP regression.

This is very useful information, thank you!