This paper presents statistical methods used in an evaluation of the Presidential Youth Fitness Program (PYFP), a multi-component intervention designed to improve standard physical education (PE) practices in schools. The evaluation was supported through a partnership between the National Fitness Foundation and the CDC with additional funding from the Robert Wood Johnson Foundation. The evaluation used a mixed-methods, two-group, quasi-experimental design with more than 2,000 students in 26 middle schools (13 PYFP and 13 comparison) with at least 4 classes per school. Data sources included FitnessGram® assessments, accelerometry recordings, and surveys with students, PE teachers and administrators. PYFP and comparison schools were matched on school size, percentage of free and reduced lunch students, geographic area, and racial/ethnic distribution. For each outcome measure, we developed multilevel multivariate models that incorporated teacher and school effects and controlled for a range of demographic covariates. This analysis is focused on the objective measures of physical activity for more than 500 students during one week twice in the semester.