Page 1 of 1

Interpreting results of a partial mediation

Posted: Tue Mar 07, 2006 9:49 am
by saab
Hi all,

this is a somewhat basic question:

I have a causal model with 2 endogenous (X, Y) and 3 exogenous (A, B, C) LV's. A and B are antecedents of X (R-Square 50%). C together with X are antecedents of Y (R-Square 60 %).
In this case, there are only direct effects from C and X on Y (no arrow between C and X).
When modeling a partial mediation (i.e. indirect effect from C to Y through X) I get the following results: R-Square of X increases (positive f-square) but R-Square of Y decreases (negative f-square).

How can this result be interpreted adequately, since I can not estimate a "total effect"? Is this a kind of "method factor", because there is an "additional path"?

Thanks for any comments,
Samy