Purpose: Preventing obesity in childhood is an increasingly important public health goal. Prevention efforts can be improved by better understanding relationships between health behaviors and overweight and obesity. This study examined such relationships in young American Indian and white children living in the rural United States. Methods: Self‐report measures of diet, screen time (passive and active), and physical activity were combined with cardiovascular fitness in cross‐sectional analyses to predict weight categories based on body mass index percentiles in 306 American Indian and white children (aged 8‐9 years) from a rural area in the upper Midwestern United States. Findings: Multinomial logistic regression models were statistically significant for girls (χ2[20]= 42.73, P < .01), boys (χ2[20]= 50.44, P < .001), American Indian (χ2[20]= 36.67, P < .05), and white children (χ2[20]= 55.99, P < .001). Obesity was associated with poorer cardiovascular fitness in girls (OR = 0.82), boys (OR = 0.83), American Indian (OR = 0.79), and white children (OR = 0.85), and with passive screen time in girls (OR = 1.69), boys (OR = 2.1), and white children (OR = 1.81). Overweight was associated with passive screen time (OR = 2.24) and inversely with active screen time (OR = 0.54), but only in boys. Conclusions: Logistic regression models were more successful at predicting obesity than overweight in all groups of participants. Poorer cardiovascular fitness showed the strongest and most consistent association with obesity, but passive screen time was also a significant and important contributor to the prediction of obesity in most prediction models. Prediction models were similar in girls, boys, American Indian, and white children.