Table 1.
Numeric results for different models
| Analysis | VPA | MPEP | VPA × MPEP | GOF | AIC | P model |
|---|---|---|---|---|---|---|
| Normal | 0.055 | 0.009 | 0.178 | <0.001 | 259 | <0.001 |
| Poisson | <0.001 | <0.001 | 0.223 | <0.001 | 256 | <0.001 |
| Quasi-Poisson | 0.045 | 0.006 | 0.490 | <0.001 | NA | NA |
| Negative binomial | 0.075 | 0.009 | 0.470 | 0.233 | 223 | 1.000 |
| Wilcoxon | ||||||
| All: 0.602 | All: 0.002 | NA | NA | NA | NA | |
| MPEP: 0.494 | VPA: 0.021 | NA | NA | NA | NA | |
| Vehicle: 0.138 | Control: 0.044 | NA | NA | NA | NA |
P-values for main and interaction effects are displayed. Goodness-of-fit p-values (a small p-value indicates a poor fit), AIC values (lower is better), and model probabilities (higher is better) indicate that the negative binomial model is preferred. The nonparametric Wilcoxon tests give much larger p-values overall.