We performed statistical analyses using two programs. We first examined gender and racial/ethnic differences on key variables of interest in SPSS 18.0, and then used structural equation modeling (SEM) in Mplus Version 5.0 to examine the proposed mediation model. Prior to analyses, we mean-centered all predictor variables to aid in the interpretation of findings (Aiken & West, 1991). We estimated SEMs with full information estimation (FIE), which uses all available data to estimate parameters and does not exclude observations with missing data (Kline, 2005). When compared to pairwise and listwise deletion, FIE is more efficient and less biased for handling missing data (Arbuckle, 1996). In addition, we used the bias-corrected bootstrap method to test the significance of mediated paths. As detailed by MacKinnon, Lockwood, and Williams (2004), the bias-corrected bootstrap method provides a more favorable balance between Type 1 and 2 errors than do other methods used to test the significance of mediated paths. Altogether, we used 500 bootstrap samples and 95 % bias-corrected confidence intervals (CIs) to estimate the significance of the mediated paths.
