TABLE 2—
Nested Cross-Classified Multilevel Models Showing the Associations Between Predictors and Depressive Symptoms: National Longitudinal Study of Adolescent Health, 1994–1995
| Model 3 (School-Level Predictors) | Model 4 (Neighborhood-Level Predictors) | Model 5 (School- and Neighborhood-Level Predictors) | |
| Fixed-effect estimates | |||
| Intercept (SE) | 3.91* (0.8) | 4.01* (0.8) | 3.98* (0.9)* |
| Individual-level predictor | |||
| Age | 0.40* (0.33, 0.48) | 0.39* (0.32, 0.47) | 0.40* (0.32, 0.48) |
| Female | 1.97* (1.74, 2.20) | 1.97* (1.74, 2.20) | 1.97* (1.75, 2.20) |
| Public assistance | 1.63* (1.23, 2.04) | 1.60* (1.19, 2.00) | 1.57* (1.17, 1.97) |
| College degree (parent) | −1.41* (−1.68, −1.14) | −1.37* (−1.64, −1.09) | −1.36* (−1.62, −1.09) |
| White (Ref) | 1.00 | 1.00 | 1.00 |
| Black | 0.84* (0.46, 1.22) | 0.72* (0.30, 1.13) | 0.68* (0.27, 1.09) |
| Hispanic | 1.37* (0.95, 1.79) | 1.41* (1.02, 1.81) | 1.33* (0.92, 1.74) |
| School-level predictor | |||
| Public assistance, % | 0.02 (−0.02, 0.05) | … | 0.01 (−0.03, 0.04) |
| College degree, % | −0.01 (−0.02, 0.01) | … | 0.00 (−0.02, 0.02) |
| White, % | −0.01 (−0.02, 0.00) | … | −0.01 (−0.02, 0.01) |
| Neighborhood-level predictor | |||
| Public assistance, % | … | 0.02 (0.00, 0.05) | 0.02 (−0.01, 0.05) |
| College degree, % | … | −0.01 (−0.02, 0.00) | −0.01 (−0.03, 0.00) |
| White, % | … | 0.00 (0.00, 0.01) | 0.00 (−0.01, 0.01) |
| Random-effect estimates | |||
| Neighborhood (95% CI) | 0.20* (−0.02, 0.42) | 0.15 (−0.03, 0.33) | 0.15 (−0.02, 0.31) |
| School (95% CI) | 0.65* (0.33, 0.96) | 0.73* (0.40, 1.05) | 0.73* (0.38, 1.08) |
| Individual (95% CI) | 52.82* (51.6, 54.0) | 52.80* (51.62, 53.98) | 52.80* (51.63, 53.98) |
| Fit statistics | |||
| Deviance information criteriona | 110 168 | 110 157 | 110 158 |
| Intraclass correlation coefficientb estimates, % | |||
| Neighborhood | 0.4 | 0.3 | 0.3 |
| School | 1.2 | 1.4 | 1.4 |
Note. CI = credible interval. The sample size was n = 16 172. For fixed-effect estimates, entries are parameter estimates (b) and credible intervals. All other entries are estimates (variance) and credible intervals.
a
A measure of model fit reported only for cross-classified multilevel models. Higher values indicate a poorer fitting model.
b
Proportion of variance in the outcome attributable to neighborhoods (after adjustment for schools) or to schools (after adjustment for neighborhoods).
*P < .05.