Thus, multilevel growth models disaggregate variance in healthy behaviors into individuals’ attributes that are stable over time and individuals’ attributes that change over time, while controlling for the clustering of Add Health respondents in schools at wave I. The benefit of this approach is that I am able to evaluate the relationships between psychosocial, social support, and family of origin resources and adolescent healthy behaviors at age 13 – while adolescents are living at home with parents – as well as the relationships between these variables and adolescents’ rate of change in healthy behaviors over time, between ages 13 and 24. I restructure the three waves of Add Health data so that the time metric is years of age, rather than waves of survey data (see Costello, Swendsen, Rose, & Dierker, 2008 for a similar approach). As a result, I estimate the average rate of healthy behavior change per year from age 13 (the youngest age in the sample at wave I) to age 24 (the oldest age for which there is an adequate sample size at wave III).