Table 1.
Leave-one-out model comparison results suggest the full model, including amygdala–mPFC functional connectivity and OXTRm fits best.
| Model | ELPD difference | ELPD SE | Bayesian R2 | Lower 95% CI | Upper 95% CI |
|---|---|---|---|---|---|
| Full model | 0.00 | 0.00 | 0.41 | 0.22 | 0.55 |
| No functional connectivity or OXTRm | −3.28 | 4.84 | 0.10 | 0.02 | 0.24 |
| No functional connectivity x OXTRm interaction | −4.89 | 4.57 | 0.14 | 0.03 | 0.30 |
| No functional connectivity | −5.00 | 4.77 | 0.13 | 0.03 | 0.28 |
| No OXTRm | −6.31 | 3.97 | 0.24 | 0.06 | 0.42 |
The expected log pointwise predictive density (ELPD) estimates the generalizability of the model for new data, and the difference is relative to the best model. Models are ranked from top (better) to bottom (worse) in terms of fit.