This study aimed to validate body composition analysis using bioelectrical impedance analysis (BIA) against dual‐energy X‐ray absorptiometry (DXA) in children with obesity and to compare agreement between BIA and DXA according to their degree of obesity. Three hundred and sixteen children aged 6‐17 years participated in the Intervention for Childhood and Adolescents Obesity via Activity and Nutrition study. We divided participants by body mass index (BMI) percentile (group 1: mild to moderate obesity; group 2: severe obesity) and compared body composition variables, eg, percentage of body fat (%BF), fat mass (FM), and fat‐free mass (FFM) using BIA and DXA. The %BF and FM of BIA were significantly lower (−1.8% and −0.8 kg, respectively), and the FFM of BIA was significantly higher (1.4 kg) than those of DXA. There were significant negative relationships between the absolute value of differences from BIA and DXA and BMI z‐scores in %BF, FM, and FFM (regression coefficient β: −1.39, 95% confidence interval CI: −1.81 to −0.97; β: −0.34, 95%CI: −0.61 to −0.06; β: −0.73, 95%CI: −1.03 to −0.44, respectively). The gap of body compositions between BIA and DXA decreased as participants became more obese, and the differences of FM in boys with severe obesity and the differences of %BF and FFM in girls with severe obesity were much less than those in children with mild to moderate obesity. In conclusion, the agreement between DXA and BIA was better for children with severe obesity than for children with mild to moderate obesity.