Interpretation | Linear model | Poly2 | Hill’sEq | F&M’sEq | |
---|---|---|---|---|---|
SSE (N2) | Sum of squared errors | 40,581 ± 32 443 [8111–100 708] | 14,395 ± 13,278 [1150–42547] | 14 507 ± 13,311 [1980–41599] | 15,380 ± 14,256 [1846–41855] |
AICc | Index of information lost by approximating the observed data (Kullback–Leibler estimate). Lower values represent less lost information and proximity to “reality” | 69.054 ± 5.244 [61.255–76.369] | 91.884 ± 6.825 [79.535–101.200] | 92.031 ± 6.487 [82.793–101.064] | 92.188 ± 6.754 [82.373–101.101] |
ΔAICc | Difference to the best model AICc. Allows for model ranking and assessing relative performance | Ø | 22.830 ± 4.822 [13.953–28.738] | 22.977 ± 5.302 [11.124–28.682] | 23.134 ± 5.457 [10.703–28.794] |
AICcw | Relative weight of evidence for each model, as the probability for being the best model for the observed data, given the candidate set of models | 0.999 ± 0.003 [0.991–1.000] | 0.000 ± 0.000 [0.000–0.001] | 0.000 ± 0.001 [0.000–0.004] | 0.001 ± 0.002 [0.000–0.005] |
AICcw-ER | Quantification of the strength of evidence in favor of best model. Practically, “how much less likely the model is than the best model?” | Ø | 435,608 ± 611,109 [1071–1739075] | 446,844 ± 599,662 [260–1690866] | 472,653 ± 621,639 [211–1788623] |
AICcw-ER (%) | Ratio of AICcw of the compared model to the AICcw of the best model, corresponding to a normalized probability that the best model is to be preferred | Ø | 99.988 ± 0.031 [99.907–100.000] | 99.956 ± 0.127 [99.617–100.000] | 99.947 ± 0.157 [99.528–100.000] |