Table 3 Multinomial logistic regression models: correlates of reported body weight control practices among urban, low-income African American students*

Correlates All† Boys‡ Girls‡
OR (95% CI) OR (95% CI) OR (95% CI)
Body weight status and body weight perception
Y = Try to lose weight Model 1 Model 2 Model 3
BMI ≥ 85th 4.52 (2.53, 8.08) 3.17 (1.45, 6.89) 7.22 (2.93, 17.81)
Perceived overweight 18.04 (7.19, 45.30) 5.88 (1.99, 17.33) 95.0 (11.91, 757.9)
Perceived underweight 1.58 (0.53, 4.70) 0.88 (0.21, 3.78) 3.30 (0.62, 17.57)
Feel bad about weight 4.12 (1.64, 10.37) 4.85 (1.22, 19.24) 2.98 (0.72, 12.34)
Y = Try to gain weight
BMI ≥ 85th 0.40 (0.13, 1.22) 0.50 (0.13, 1.87) 0.21 (0.02, 2.02)
Perceived overweight 0.0002 (0.0001, 0.0003) 0.0002 (0.0001, 0.0004) 0.002 (0.001, 0.005)
Perceived underweight 12.36 (5.57, 27.44) 8.99 (3.11, 25.96) 18.86 (5.42, 65.67)
Feel bad about weight 5.86 (2.09, 16.43) 7.01 (1.42, 34.47) 4.36 (1.09, 17.48)
Body weight status and misclassification
Y = Try to lose weight Model 4 Model 5 Model 6
BMI ≥ 85th 26.33 (12.36, 56.08) 9.18 (3.64, 23.15) 112.7 (24.41, 519.9)
Overestimate weight status 14.47 (3.88, 53.92) 12.02 (1.32, 109.1) 17.62 (3.35, 92.77)
Underestimate weight status 0.21 (0.10, 0.47) 0.46 (0.18, 1.20) 0.06 (0.01, 0.30)
Feel bad about one's weight 4.81 (1.99, 11.60) 4.39 (1.10, 17.55) 4.03 (1.14, 14.22)
Y = Try to gain weight
BMI ≥ 85th 0.08 (0.02, 0.25) 0.12 (0.03, 0.47) 0.03 (0.003, 0.29)
Overestimate weight status 1.23 (0.18, 8.48) 2.56 (0.29, 22.74)
Underestimate weight status 5.92 (2.82, 12.40) 3.94 (1.45, 10.69) 10.60 (3.33, 33.76)
Feel bad about one's weight 7.84 (2.92, 21.03) 8.93 (1.97, 40.47) 5.33 (1.37, 20.70)
1. *Participants whose self-reported weight control practice was 'stay the same' were treated as the reference group for the outcome. The following participants were treated as references groups in the corresponding models for independent variables: perceived weight = 'normal'; feel bad about one's weight = 'no'; BMI = ' < 85th percentile'; comparison of weight status between self-evaluation and that based on measured BMI = 'correspondence'.
2. † Gender and grade were controlled in each model; ‡ Grade was controlled in each model.
3. §Could not be estimated because no boy who overestimated their body weight status reported trying to gain weight. To check model fit, we had re-fit all these models by artificially assigned a boy who overestimated their body weight status to try to gain weight. All the results were consistent except for the correlates of 'overestimate weight status.'