. . . . "2026-05-29"^^ . . "Target-group background improves MaxEnt AUC — Phillips Table 2 reproduced" . "The reproduction confirms Phillips et al. 2009 Table 2: target-group background improves the mean predictive AUC of presence-only MaxEnt models. Across 225 species in six regions, mean AUC rises from 0.7163 with random background to 0.7468 with target-group background, a gain of +0.0305. This closely matches Phillips' reported Maxent gain of +0.0293 (0.7276 to 0.7569) in both direction and magnitude, despite using the independent open-source elapid/maxnet engine rather than the original Java Maxent. Phillips' secondary finding that the gain is largest where sampling bias is greatest also reproduces: the Canadian region (CAN), the most biased, shows the largest gain." . . "Overall mean AUC: random 0.7163, target-group 0.7468, delta +0.0305 (Phillips Maxent: 0.7276, 0.7569, delta +0.0293). Paired Wilcoxon signed-rank test of target-group vs random per-species AUC: statistic 8028.0, p = 1.6e-06 (significant at p < 0.001). Per-region deltas: CAN +0.1317 (largest, 20 species), SWI +0.0394 (30), AWT +0.0451 (40), NSW +0.0184 (53), NZ +0.0135 (52), SA -0.0146 (30, the only region where target-group background did not help). The largest gain falls on CAN, the region flagged as most sampling-biased, reproducing Phillips' bias-gradient result.\n\nGithub: https://github.com/annefou/sdm-phillips-reproduction" . "(1) The MaxEnt engine differs from Phillips' original (elapid/maxnet vs Java Maxent), so absolute AUC decimals differ even though direction and magnitude agree.\n(2) Only MaxEnt is reproduced; Phillips' broader Table 2 also covered BRT, MARS, GAM and other methods, so the cross-method generality of the claim is not tested here. \n(3) 225 of Phillips' 226 species are modelled — one species was dropped because its presence-absence evaluation column had no presence/absence variation, leaving AUC undefined. \n(4) A minimum-presence threshold of 5 occurrences was applied. \n(5) One region (SA) showed a small negative gain, consistent with the original's observation that the benefit is not uniform across regions." . . . . . "Anne Fouilloux" . "2026-05-31T08:19:45.828Z"^^ . . . . . "Target-group background improves MaxEnt AUC — Phillips Table 2 reproduced" . . "RSA" . "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" . "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" . . .