Table 2.
Intervention marginal effects on BD: two-part model (logit + glm)
| Marginal effects | ||
|---|---|---|
| Intervention | ||
| Period | 0.279 (0.20) | |
| Treated | 0.201 (0.14) | |
| Socioeconomic | ||
| Age | 0.265 (0.05)*** | |
| Female | 0.068 (0.09) | |
| Spanish | 0.226 (0.25) | |
| Nuclear family composition | −0.110 (0.08) | |
| Mother’s schooling years | − 0.002 (0.00)* | |
| Good economic situation at home | 0.030 (0.09) | |
| Pocket money (weekly) | 0.022 (0.01)*** | |
| Completed questionnaire more days after last weekend | − 0.330 (0.11)*** | |
| Completed questionnaire near to local events | − 0.001 (0.00) | |
| Alcohol | ||
| Family alcohol consumption | 0.502 (0.06)*** | |
| N x T | 1638 | |
| Wald χ2 | 182.72 (0.00) | |
| Log likelihood | − 1432.04 | |
| Pseudo-R2 1st part | 0.0774 | |
Note: we show the parameter estimates and standard deviations in brackets. ***, ** and * represent statistical significance at 1, 5 and 10%; 1000 replications were used for bootstrapping and standard errors were clustered at classroom level. In the two-part model and the finite mixture model, regressions were also controlled by period of intervention, female sex, age, Spanish nationality, having a partner, years of schooling of the mother, nuclear family, pocket money, family pressure, answering the questionnaire late in the week and answering the questionnaire near the date of local events