*2.3. Data Analytics Plan*

First, to rule out possible common method variance bias due to the self-report nature of our data [29,30], we applied Harman's single-factor test on all measures prior to data analysis [31]. Based on an eight-factor solution explaining 65.5% of variance, with the first factor accounting for 36.4%, the bias of common method variance in the present data was assumed to be low [30]. Using univariate analyses of covariance (ANCOVA; IBM® SPSS version 25.0; Chicago, IL, USA) on the MOB-K sum score by weight status, sex, and their interaction, weight- and sex-specific differences were identified, controlling for age and SES. Although all dependent variables deviated from normal distribution, which is common in large samples [32], ANCOVA was used because of its robustness against non-normality [33] and the large sample being selected with a high degree of randomization. However, all analyses were repeated using non-parametric methods, specifically the Kruskal–Wallis test, and their results will be reported if deviating from parametric analyses. Post-hoc analyses included unpaired t-tests, applying Bonferroni-corrected significance levels. Cohen's *d* was used as an effect size measure, representing small (≥0.2), medium (≥0.5), and large (≥0.8) effects [34].

Second, using path analysis (IBM® SPSS AMOS® version 25.0; Chicago, IL, USA), we tested the possible mediating effect of experiences of workplace bullying on the association between weight status and psychological health impairments. As a first step, direct effects of weight status on psychological health impairments (BOSS II and EQ-5D) were examined for the total sample (Model 1), controlling for socioeconomic variables (sex, age, and SES). Subsequently, the indirect effects of weight status on mental health variables, possibly mediated by the experiences of workplace bullying (MOB-K sum score), were examined (Model 2).

Third, possible sex-specific mediation effects (i.e., moderated mediation [35,36]) of the experiences of workplace bullying on the associations between weight status and psychological health impairments were examined using the SPSS PROCESS macro version 3.5 [36], which utilizes bootstrapping to assess direct and indirect effects of variables while maximizing power and minimizing concerns about non-normality. Furthermore, 95% confidence intervals were resampled 5000 times for each analysis to test the significance of the indirect effects [37]. To additionally confirm and depict sex-specific differences regarding the explanatory power of weight status and experiences of workplace bullying for psychological health impairments, the final model from the path analysis (Model 2) was applied to subsamples of women (Model 3) and men (Model 4) separately. For path analyses, the following indices were determined for the evaluation of model fit: χ<sup>2</sup> test statistics; the minimum discrepancy, divided by its degrees of freedom (CMIN/DF); the comparative-fit index (CFI); the Tucker–Lewis index (TLI); the normed-fit index (NFI); and the root mean square error of approximation (RMSEA). Good model fit is indicated by nonsignificant χ<sup>2</sup> statistics; CMIN/DF < 2; CFI, TLI, and NFI > 0.97; and RMSEA < 0.08 [38]. Standardized regression weights were interpreted as indicative of small (≤0.30), medium (between 0.30 and 0.50), or large (>0.50) effects [34]. All analyses used a two-tailed α < 0.05 as the significance level.
