Table 1.
Model performance – quantified using AUC, accuracy and TSS – declined when the full model was used to predict on helminth taxonomic subsets when considering the global set of interactions between hosts and helminth parasites
| Performance | Helminth taxa | t | df | P value |
|---|---|---|---|---|
| AUC | Acanthcephalans | 4.65 | 92.4 | 0.0001 |
| Platyhelminthes | 0.76 | 97.8 | 0.45 | |
| Nematodes | 1.79 | 97.5 | 0.077 | |
| Accuracy | Acanthcephalans | 10.44 | 94.5 | 0.0001 |
| Platyhelminthes | 2.96 | 94.9 | 0.003 | |
| Nematodes | 2.00 | 96.4 | 0.049 | |
| TSS | Acanthcephalans | 5.44 | 97.5 | 0.0001 |
| Platyhelminthes | 1.21 | 97.1 | 0.229 | |
| Nematodes | 2.28 | 97.8 | 0.025 |
Model performance was compared to helminth group subset predictions from the full model using Welch's two-sample t-tests across the 50 trained boosted regression models. Bold P-values indicate significance assessed at α = 0.05.