Figure 2.

Cross‐validation of conventional kernel density estimation (KDE) across the mammalian body‐mass spectrum: (a) percentage of locations from the second half of the data (held‐out set) included in KDE 50% home ranges estimated from the first half of the data (training set) as a function of body mass (dashed line, target 50% quantile; solid line, phylogenetically controlled regression model fit to cross‐validation results; shading, 95% CI of the fit) and (b) regression model describing the accuracy of 95% KDE area estimates across the mass spectrum. Accuracy was quantified as the ratio between estimated 95% area of the training set and the area contained within the contour that encompassed 95% of locations in the held‐out set. The horizontal dashed line represents an unbiased area estimate. The x‐axes in are log scaled.