Fig. 4

Directed acyclic graph (DAG) of the known or assumed associations between variables of the illustrative example. For the sake of simplicity and clarity, the arrows representing the associations between the covariates are not drawn. Not all types of variables considered in the simulation study were suitable for this illustrated example. The covariates ‘maternal educational level’ and ‘maternal place of birth’ were considered to be confounding factors in the relationship between pre-pregnancy maternal BMI and child BMI at age 7. The association models were therefore adjusted for these covariates