Table 2.
Multilevel logistic models of obesity.
| Model 1 | Model 2 | Model 3 | Model 4 | |
|---|---|---|---|---|
| Fixed effects | ||||
| Intercept | −1.966*** (.132) | −4.882*** (.299) | −8.893*** (1.195) | −9.459*** (1.284) |
| Individual-level variables | ||||
| Income (in thousands) | -.0001 (.0002) | -.0001 (.0001) | -.0002 (.0002) | |
| Gender (male) | .032 (.031) | .032 (.031) | .032 (.031) | |
| Education | -.052*** (.004) | -.052*** (.004) | -.052*** (.004) | |
| Age | .142*** (.006) | .142*** (.006) | .142*** (.006) | |
| Age2 | -.0013*** (.0001) | -.0013*** (.0001) | -.0013*** (.0001) | |
| Residence (rural) | .082* (.033) | .080* (.033) | .080* (.033) | |
| Working status | ||||
| Currently working | Reference | Reference | Reference | |
| Formerly worked | .374*** (.037) | .374*** (.037) | .376*** (.037) | |
| Never worked | .308*** (.066) | .310*** (.066) | .312*** (.066) | |
| Marital status | ||||
| Single | Reference | Reference | Reference | |
| Married | -.072 (.046) | -.071 (.046) | -.071 (.046) | |
| Divorced | -.288*** (.063) | -.286*** (.063) | -.286*** (.063) | |
| Widowed | .177* (.068) | .181** (.068) | .181** (.068) | |
| Country-level variables | ||||
| GDP per capita (in thousands) | -.009 (.013) | -.012 (.013) | ||
| Globalization | .054** (.018) | .065*** (.018) | ||
| Random effects (variance across countries) | ||||
| Intercept | Included | Included | Included | Included |
| Slope (income) | Included | |||
| LR test | 1756.6*** | 1889.7*** | 1361.4*** | 1361.8*** |
| Model fit statistics | ||||
| Wald χ2 | – | 1062.7*** | 1075.0*** | 1100.8*** |
| AIC | 32719.6 | 31082.7 | 31075.1 | 31074.7 |
| BIC | 32737.0 | 31195.7 | 31189.5 | 31185.1 |
Notes: (1) Numbers in parentheses are standard errors; (2) from 2-tailed tests, *P < .05; **P < .01; ***P < .001; (3) under “Random Effects” the LR test reports the likelihood ratio statistic for testing the null hypothesis that the random effects (between-country variance) are zero. A significant test result indicates that it is necessary to include random effects in the modeling.