Table 4.
Selection table of the fits of the GLS models of the association between assemblage-level body mass and assemblage-level NDVI values
| ln L | AIC | ΔAIC | wi | |
|---|---|---|---|---|
| No spatial autocorrelation structure | −7504.1 | 15014.2 | 13952.2 | <0.001 |
| Exponential autocorrelation (Euclidean) | −726.9 | 1463.7 | 401.7 | <0.001 |
| Exponential autocorrelation (haversine) | −526.0 | 1062.0 | 0.0 | 0.988 |
| Gaussian autocorrelation (Euclidean) | −1046.0 | 2102.0 | 1040.0 | <0.001 |
| Gaussian autocorrelation (haversine) | −962.3 | 1934.6 | 872.5 | <0.001 |
| Linear autocorrelation (Euclidean) | −1832.7 | 3675.5 | 2613.5 | <0.001 |
| Linear autocorrelation (haversine) | −1347.3 | 2704.7 | 1642.6 | <0.001 |
| Rational quadratics autocorrelation (Euclidean) | −808.2 | 1626.4 | 564.4 | <0.001 |
| Rational quadratics autocorrelation (haversine) | −682.2 | 1374.4 | 312.3 | <0.001 |
| Spherical autocorrelation (Euclidean) | −733.2 | 1476.4 | 414.3 | <0.001 |
| Spherical autocorrelation (haversine) | −530.4 | 1070.8 | 8.8 | 0.012 |
ln L, restricted log-likelihood score; ΔAIC fit relative to the model with the lowest AIC score (italicized). The best-fit model based on ΔAIC and Akaike weights (wi) are denoted in bold. The degrees of freedom (n − 2) for all models is 7,600 (n − 2).