FIGURE 3. Results from the machine-learning Linear Discriminant Analysis (LDA) classifier of phenotypic limits in Trachops. Overall accuracy of the model = 91.8% (95% CI 89.7%–93.6%). A. In biplot, percent group separation described for the first linear discriminant (LD1) was 96, and 4 for the second linear discriminant (LD2). Solid lines represent 68% data ellipses centered at the bivariate mean to visualize phenotypic differences among species. LD1 density values plotted for aid in visualization on the x-axis highlighting closer phenotypic affinities of coffini (red) and ehrhardti (purple) compared to cirrhosus (gray, all five geographic groups included). B. Map of the distribution of individual specimens measured for morphological analyses.