Table 1. Comparative performance of models.
Model | Params | AIC | Links | Entries | |
PNM | 87 | −105 | 385 | 0.797 | 0.906 |
n, c free; r = f(c) | 60 | −141 | 402 | 0.734 | 0.883 |
n, c free; r = f(n) | 60 | −151 | 421 | 0.734 | 0.880 |
n, r free, c = f(n) | 60 | −174 | 469 | 0.754 | 0.875 |
n free, c = f(n), r = g(n) | 33 | −205 | 475 | 0.712 | 0.855 |
ANM | 58 | −209 | 535 | 0.681 | 0.845 |
MPNM | 32 | −214 | 493 |
Params is the number of model parameters; is maximum log-likelihood; Links is the fraction of links reproduced by the model; Entries is the expected fraction of connection matrix entries reproduced by the model. PNM is probabilistic niche model with ni, ci and ri all free parameters. ANM is the allometric niche model. MPNM is the minimum potential niche model [18]. Other models are variants of the PNM with one or two parameters a linear function of another parameter.