Fig. 1

Effects of selected variables on the percentiles of the BMI distribution estimated through multivariate quantile models. On the horizontal axes are the BMI distribution percentiles; the vertical axes show the values of the coefficients estimated. Dashed parallel line represents the null value (zero), and a solid line stands for quantile estimates. A grey area surrounding the solid line represents the 95 % confidence intervals (CI) for the quantile estimates. The explanatory variables evaluated were: age, domestic overload, years worked at night, self-rated health, physical inactivity, and BMI at age 20 years. All coefficients are adjusted by the other study variables