Fig. 1
From: Artificial intelligence reveals environmental constraints on colour diversity in insects

Comparison between the predicted and actual mean elevations of moth species. a There were 1951 and 1047 species in the complete and validation datasets, respectively. b For a subset of the data with >10 images in the training dataset, there were 500 and 480 species in the training and validation dataset, respectively. c For the subset of species with ≤10 images in the training dataset, there were 1451 and 567 species in the training and validation dataset, respectively. d For the subset of species with larger elevational distribution ranges (range size larger than the median size), there were 976 and 745 species in the training and validation dataset, respectively. e For the subset of species with smaller elevational distribution ranges (range size smaller than the median size), there were 975 and 302 species in the training and validation dataset, respectively. The results of a GLMM controlling for phylogenetic effects are shown at the bottom of each figure. Source data are provided as a Source Data file