Fig. 3

Plots of the proportion of individual elk whose parturition status was correctly identified (accuracy). Overall classification accuracy (a) combines correct classification of parturient elk used to build the random forest model (n  = 37) and elk that were either non-parturient or lost their calf within the first week post-parturition, used to test the model (n  = 20). b Sensitivity and only includes parturient elk from the training set and c specificity and only includes non-parturient/calf loss elk from the testing set. Gray lines indicate the chosen decision probability threshold (0.75). Each color/line type indicates the number of days at or above the probability threshold that we considered