The findings are consistent with the hypothesis that non-random selection into and
out of neighbourhoods accounts for some of the association between BMI and neighbourhood
characteristics. Future studies may profit by adopting the current methodology to assess the
amount of influence estimated due to residential selection