KDA posted on Thursday, December 29, 2005 - 10:42 am
Dear Linda and Bengt:
I have specified a SEM with the following paths: 1) direct paths from observed variables A, C to observed variable B. 2) In addition to (1) above, I have also specified a path from C to latent construct D, and from D to B.
I have two questions I'd like to ask about this model:
1. How do I test the hypothesis that the mean level of B exceeds that of A, controlling for all the paths specified above? 2. Is it correct to compare the direct path from C to B with the product of the paths from C to D and D to B (joint effect) if I am interested in the relative strength of the direct versus mediated path to B?
I would greatly appreciate your advice.
bmuthen posted on Thursday, December 29, 2005 - 6:15 pm
1. Seems like you want to test the intercept for B.
I was focusing on the phrase "controlling for" in 1., which to me suggests looking at differences conditioning on something. That led me to suggest focusing on the intercept - in line with the ANCOVA answer I gave you.
But perhaps I misunderstood, and the individual who asked is more interested in growth matters.