Childhood obesity rates in the United States increased dramatically between the 1980’s and the 2000’s and remain high in all demographic sub-groups. The rates are particularly high in African American and Hispanic children. In response to these trends and following the recommendations of leading health authorities, many U.S. communities have implemented programs and policies aimed at improving children’s diets, increasing their physical activity levels, and reducing childhood obesity. However, the impact of these efforts has been limited, and neither research nor professional practice has yet identified the attributes of community programs and policies that are consistently successful in improving children’s diet, increasing their physical activity, and decreasing the probability that they will develop overweight or obesity. The overarching goal of the proposed research is to produce findings that inform public health policies and practices aimed at reducing disparities in the prevalence of childhood obesity. Specific aims will be to identify attributes of community programs and policies that are associated with children’s diet, physical activity and weight status within groups of communities, schools, and families stratified by race/ethnicity and socioeconomic status. These aims will be addressed by conducting secondary analyses in an existing dataset that includes in-depth observations of a diverse sample of over 5000 children and assessment of over 9500 community programs and policies to reduce childhood obesity in 130 U.S. communities. Further, within those communities, extensive data on school nutrition and physical activity policies, practices and resources are available for 274 elementary and 149 middle schools. A home-based data collection protocol provided anthropometric data as well as detailed information on children’s diet and physical activity behaviors. Two analytic methods, Classification and Regression Trees (CART) and Gradient Boosting Machines (GBM), will be employed in examining relationships between attributes of community programs and policies and children’s diet, physical activity and weight status. These analytic procedures will be conducted within sub-groups of communities, schools and families stratified by race/ethnicity and socioeconomic status profiles. The findings of the proposed study will enable researchers and practitioners to tailor community-based interventions to the characteristics of demographic sub-groups. This research will be significant because it will inform strategies for preventing childhood obesity in communities and population sub-groups that are disproportionately affected by the negative consequences of obesity. The study will be innovative because it will apply state-of-the-art analytical methods using a dataset that is uniquely large and comprehensive.