Statistical changes have been made to account for study attrition to permit for generalizations to the baseline school-based sample and to help modify for possible choice bias. In all analyses, information had been weighted using a response propensity method. Response propensities (i.e., the likelihood of responding to the EAT-IV survey) had been estimated using a logistic regression of response at 15-year follow-up on predictor variables from the school-based survey. Study variables had a varying amount of missing data (0–13%) and all analyses were performed with complete circumstances. The pattern size for particular person variables is proven in Table1.

Leave a Reply