Table 2. Regressional mediation analyses showing the direct effect of gender (1 = boy, 2 = girl) on educational functioning and educational risk, and the indirect effect of gender through exposure to community violence and harsh physical discipline. All direct effects and indirect effects were significant.
| Educational functioning score | Number of educational risks | |
|---|---|---|
| B (CI) | B (CI) | |
| Gender | 0.61 (0.27, 0.95) | −0.19 (−0.31, −0.064) |
| Community violence | −0.18 (−0.39, 0.021) | 0.082 (0.0071, 0.16) |
| Harsh physical discipline | −0.33 (−0.58, −0.06) | 0.16 (0.066, 0.26) |
| Carer HIV status | −0.032 (−0.48, 0.42) | 0.082 (−0.083, 0.25) |
| Digit span | 0.059 (0.013, 0.10) | −0.062 (−0.078, −0.045) |
| Draw-a-person score | −0.0086 (−0.018, 0.0009) | −0.0018 (−0.0053, 0.0017) |
| Any developmental delay | −1.48 (−1.83, −1.13) | 0.57 (0.45, 0.70) |
| School attendance | −1.20 (−1.92, −0.48) | 1.35 (1.09, 1.62) |
| Direct effect | 0.61 (0.27, 0.95) | −0.19 (−0.31, −0.064) |
| Indirect effect: community violence | 0.032 (0.0013, 0.096) | −0.014 (−0.040, −0.0003) |
| Indirect effect: harsh physical discipline | 0.037 (0.0060, 0.96) | −0.018 (−0.041, −0.0030) |
| R2 | 0.16 | 0.31 |
Notes: CI = confidence interval. Reported model fit for the logistic regression model (correct class for age) is the McFadden R2.