We checked the robustness of our findings, first by substituting the GSNI for the GII in our regression analysis (Table A5). GII and GSNI values were highly correlated (Figure A3) and, as expected, these regressions produced results very similar to those shown in Table 2. The interaction between gender social norms index and gender was significant across all four regressions models.
Second, we tested our models using a fixed-effects regression analysis in which all unmeasured country differences were dummy-coded (Table A6). Again, we found very similar results to the multilevel regression analysis.
出处：Gender Differences in Bullying Reflect Societal Gender Inequality: A Multilevel Study With Adolescents in 46 Countries