This study examines the independent and joint associations between several socioeconomic and behavioral characteristics and physical activity (PA) and inactivity prevalence among 68,288 US children aged 6-17 years. The 2003 National Survey of Children's Health was used to estimate PA prevalence. Multivariate logistic regression was used to estimate odds of activity and inactivity and adjusted prevalence, while least squares regression was used to model mean number of days of physical inactivity (PIA) in past month. The prevalence of PA varied substantially by socioeconomic and behavioral characteristics, with older, female, non-English speaking, and metropolitan children and those with lower socioeconomic status (SES) and neighborhood social capital having higher inactivity and lower activity levels. Children who watched television >or=3 h/day had 60% higher adjusted odds of PIA and 30% lower odds of PA than those who watched television <3 h/day. Children experiencing inadequate sleep during the entire week had 55% higher odds of PIA and 29% lower odds of PA than those who experienced >or=5 nights of adequate sleep during the week. Children whose both parents were physically inactive had 147% higher odds of PIA and 46% lower odds of PA than children whose parents were both physically active. Differentials in PIA by ethnicity, SES, television viewing, and parental inactivity were greater for younger than for older children. Subgroups such as older, female adolescents, children from socially disadvantaged households and neighborhoods, and those in metropolitan areas should be targeted for the promotion of regular physical activity and reduced television viewing time.