Table 9.

Performance of bioclimatic distribution models for four Bassaricyon species using vouchered specimen localities. Mean values are averages of 10 models run, each withholding 20% of data as test localities, while the Full Model AUC used all available data. The mean value for equal training sensitivity and specificity was used as a logistic threshold to create a range map predicting presence/absence.

	Localities	Mean Test AUC (stdev)	Full Model AUC	Mean Unregularized Training Gain	Mean equal training sensitivity and specificity (logistic threshold)
Bassaricyon alleni	43	0.901 (0.036)	0.939	1.85	0.302
Bassaricyon gabbii	18	0.977 (0.012)	0.993	4.09	0.222
Bassaricyon medius	31	0.952 (0.028)	0.988	3.76	0.119
Bassaricyon neblina	16	0.996 (0.002)	0.998	4.77	0.160