topic 3-way association model? in Statistical Procedures

3-way association model?

iressa131 — Sun, 17 Jun 2018 21:11:27 GMT

Hello,

I want to know if variable A influences the effect of variable B on variable C.

Would I use a saturated model or homogenous association (or something else entirely)? Clarification on the conceptual difference would be helpful! From what I understand, homogenous association is saying that B and C have a relationship and that relationship is identical at each level of A while the saturated model is saying that B and C have a relationship but that relationship does not have to be identical at each level of A? Thanks in advance for any help!!

Re: 3-way association model?

pau13rown — Mon, 18 Jun 2018 05:05:21 GMT

yes, "model y= a|b|c" (see here https://blogs.sas.com/content/iml/2016/03/07/collection-effects-pairwise-interactions.html)

bear in mind, there's little statistical power for interaction effects; there'd have to be a specific interest in the Q. As you said, an interaction is an inconsistency of the strength of the effect across the levels of the other factor

Re: 3-way association model?

StatDave — Wed, 20 Jun 2018 13:14:35 GMT

"I want to know if variable A influences the effect of variable B on variable C."

As you've stated it, this suggests that C is the response and A and B are predictors. If so, then you simply want to test if the A*B interaction is significant in this model:

model c = a b a*b;