“…In order to understand the extent to which variables at each level contribute to between‐level variations in DWCB, we started from four‐level random intercept null models, and added each set of covariates in the fixed part in sequence, with Model 1 (added age, BMI status, and pubertal maturity), Model 2 (added perceived household economic status, parental education, living structure), Model 3 (added school type, sex‐composition, and percentage of students who received school nutrition education), and Model 4 (added urbanicity) added (Austin et al, ). All multilevel models were performed based on MLwiN 2.28 (Center for Multilevel Modeling, UK) by employing Markov Chain Monte Carlo methods (Dunn et al, ). All fixed and random parts parameters were evaluated at the significance level of 0.05 (two‐sided), and the final model was determined based on the deviance information criterion (DIC) statistic (Dunn et al, ).…”