Page 1 of 1

VAF

Posted: Sun Aug 12, 2018 12:10 pm
by tinker18
Hello everyone,

I have three mediators in my model; Two of it are partial mediation and 1 full mediation. The full mediation shows a VAF value of 1.075 which is above 100%. I know that VAF is not required for full mediation but I'd like to know if I am calculating it correctly?

A= Independent variable, B = dependent variable, M = mediator.

A to M = 0.552
M to C = 0.469
A to B= -0.018

Therefore,
0.552 x 0.469 / (0.552 x 0.469 + (-0.018)) = 1.075%

However, according to a friend, we need to consider only the absolute value and disregard the negative sign. That is,
0.552 x 0.469 / (0.552 x 0.469 + 0.018) = 93.49%.

May I know which calculation is correct?

Thank you.

Re: VAF

Posted: Sun Aug 12, 2018 2:30 pm
by jmbecker
Your calculations seem to be generally correct. However, a problem with the VAF is that it does not work with negative effects (indirect or direct). Some people propose to only use the absolute effects to circumvent these problems, but the main question is: what do you want to measure with the VAF and why?