Table 3.
Summary of model comparison using leave-one-out (LOO) cross-validation
| PEP4LEP dataset | ||||
|---|---|---|---|---|
| Model | LOOIC | ELPD | ELPD difference | Standard error (ELPD difference) |
| Gamma | 1532.4 | −766.2 | −6.5 | 8.1 |
| Weibull | 1543.9 | −771.9 | −12.2 | 9.8 |
| Log-normal | 1519.5 | −759.8 | 0 | 0 |
| Global dataset | ||||
|---|---|---|---|---|
| Model | LOOIC | ELPD | ELPD difference | Standard error (ELPD difference) |
| Gamma | 738.1 | −369.0 | −6.3 | 5.5 |
| Weibull | 732.9 | −366.4 | −3.7 | 4.1 |
| Log-normal | 725.5 | −362.7 | 0 | 0 |
| Combined datasets | ||||
|---|---|---|---|---|
| Model | LOOIC | ELPD | ELPD difference | Standard error (ELPD difference) |
| Gamma | 2255.4 | −1127.7 | −3.8 | 8.9 |
| Weibull | 2261.7 | −1130.8 | −7.0 | 9.5 |
| Log-normal | 2247.7 | −1123.9 | 0 | 0 |
LOOIC is LOO information criterion; ELPD is expected log predictive density. LOOIC = −2 × ELPD and ELPD difference is the ELPD value relative to that of the best performing model, i.e., the log-normal model, which has the lowest LOOIC (or highest ELPD). The standard error is a measure of uncertainty for the ELPD difference relative to the log-normal model
ELPD Expected log predictive density, LOOIC LOO information criterion, PEP4LEP Post exposure prophylaxis for leprosy