Hi forum!
I used PLS for running a path model. However, I would like to compare model fit with the "reverse" model (i.e. instead of x -> M -> Y; run a model with Y -> M -> X). I realize this is not a test for reverse causality, but still would like to be able to compare model fit. My sample is too small (54) to run a similar analysis with SEM in Mplus for instance.
If i just look at the relevant r squares, I'm afraid this may distort the picture given that I have two independents (X's), but only one mediator and one dependent (ie the r square of the mediator and dependent may by definition be higher than the r square of each of the x's separately).
Any ideas on how to go about this?
Thank you in advance.
Kind regards,
Veroniek