Figure 4

(A) Performance comparison of the Gaussian and linear SVM classifiers as a function of the input feature vectors size. Performance is measured as the percentage of fecal samples correctly classified. Note that in this representation we do not distinguish the stressors included in the feature vector, but only the model dimensionality. Dark-solid lines illustrate the global tendency. (B) Comparison of some relevant Gaussian classifiers’ performance. Label inside the bar identifies the factors used to build the model. “8-dimensional” correspond to a SVM based on year, month, season, rainfall, habitat, moon phase, sex, and breeding condition; while “11-dimensional” to one based on the 11 environmental and individual factors analyzed in our study.