residual and unmeasured confounding could be present. However, adjust for potential socio-demographic confounders we did assess, including family structure and race/ethnicity did not have a substantive influence on our estimates of association.
The data used to identify adolescents family structure was collected in 2005 and thus may not have been an accur ate representation of adolescent family structure when the predictor and outcome measures for this analysis are were assessed. Our analyse are unable to rule out reverse causation in the observed associations. However, this limitation applies mainly to the outcome of disordered eating as reverse causation is unlikely with the other outcomes explored. All measures were collected by self-report and individuals with higher BMIs are more likely to under-report their weight and dietary intake and over-report their physical activi This reporting bias would most likely bias estimates towards the null. We calculated 42 tests and did not adjust for multiple comparisons.