Page 1 of 1

Model Selection | Different endogenous variables

Posted: Fri Oct 07, 2022 10:15 am
by acamilo
Hey there,

I am using SmartPLS to compare two established competing theories. There is theoretical evidence for both models and there is no consensus. The competing models have exactly the same variables, but the causal direction changes:

Theory 1: A -> B
Theory 2: B -> A

I'm trying to use model selection criteria (BIC/Akaike Weights and CVPAT) to understand which model fits the data best and has the higher predictive power. However, in this case, I would have to compare the BIC or the CVPAT (if focused on target constructs) of two different endogenous variables. Is this possible? If not, how could I compare these two models?

I'd be very grateful for any suggestions.