Table 3.
Accuracy of model selection under backward elimination, AIC and AICc. Letters in parentheses indicate the model codes formerly used by Yang and Swanson [4].
| Backward elimination | |||||
| Model | Heterogeneous parameters | p = 0.05 | p = 0.0001 | AIC | AICc |
| 1 (E) | κ, ω, c, π's | 100% | 96% | 91.7% | 91.7% |
| 2 (D) | κ, ω, c | 92% | 100% | 91.7% | 100.0% |
| 3 | κ, ω, π's | 61% | 67% | 38.9% | 33.3% |
| 4 | κ, ω | 94% | 100% | 77.8% | 83.3% |
| 5 | κ, c, π's | 88% | 92% | 62.5% | 66.7% |
| 6 | κ, c | 79% | 100% | 75.0% | 75.0% |
| 7 | κ, π's | 67% | 67% | 44.4% | 44.4% |
| 8 | κ | 83% | 100% | 55.6% | 55.6% |
| 9 | ω, c, π's | 71% | 83% | 58.3% | 58.3% |
| 10 | ω, c | 79% | 96% | 75.0% | 75.0% |
| 11 | ω, π's | 50% | 67% | 38.9% | 38.9% |
| 12 | ω | 89% | 100% | 66.7% | 66.7% |
| 13 (C) | c, π's | 83% | 88% | 62.5% | 58.3% |
| 14 (B) | c | 63% | 88% | 58.3% | 58.3% |
| 15 | π's | 56% | 61% | 38.9% | 38.9% |
| 16 (A) | none | 83% | 100% | 38.9% | 55.6% |
| overall | 78% | 88% | 63% | 64% | |