Logistic cubic spline regression model predicting risk of WBV across the observed range of participant BMI. Spline knots were created at every SD across the observed BMI range to capture the nonlinear relation between BMI and the risk of WBV. Outcome: “ever victimized at school” (0 = no, 1 = yes), predictor: BMI for age (z scores; mean: 10; SD: 1 in the reference population,40 sample mean = 11), N = 301.
Dependent variables in the linear regression models are z standardized and, thus, the coefficients reflect changes in the dependent variables in units of SDs for a 1-unit change in the independent variables. Data were imputed by using multiple imputation for gender (M = 20).