Mediation analyses were conducted using methods described by MacKinnon and colleagues (MacKinnon and Dwyer, 1993; MacKinnon et al., 2002). The method tests directly for the existence of a significant path from the independent variable (A) through the mediator (B) to the outcome (C) by computing a product of the standardized regression coefficients for the A-B and B-C associations. The Z score can then be compared to critical significance levels developed by MacKinnon. This joint significance test has been shown to produce optimal power and the most accurate Type I error rates in all cases compared to other methods, including the Sobell t-test of for mediation (see MacKinnon et al., 2002).
