Despite these strengths, there are a number of methodological limitations to the existing literature. Although several studies utilized prospective designs, the majority of studies are cross-sectional. The strongest test of mediation hypotheses requires multiple assessments of stressors, psychological processes, and psychopathological outcomes. Second, with rare exception (e.g., Hatzenbuehler, McLaughlin, & Nolen-Hoeksema, 2008; Hatzenbuehler, Corbin, & Fromme, 2008a), studies examining mediational hypotheses did not use established criteria for mediation, such as the causal steps strategy (Baron & Kenny, 1986) or the product-of-coefficients approach (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002). In addition, most studies indicated support for mediation through examining reductions in p-values, rather than conducting tests for mediation. Recent research has provided several significance tests for mediated effects that are associated with low bias and Type 1 error rate, accurate standard errors, and adequate power to detect small effects (MacKinnon & Fairchild, 2009), including the distribution of the product approach (MacKinnon, Fritz, Williams, & Lockwood, 2007) and bootstrapping (Preacher & Hayes, 2004; 2008). These are two methodological problems that are also evident in the general stress literature (Grant et al., 2006), but future studies that examine mediation hypotheses will need to follow these established procedures and utilize prospective designs.
