Skip to main content
. 2019 Dec 24;10(2):737–747. doi: 10.1002/ece3.5921

Figure 3.

Figure 3

(a) Model coefficients from GLM model 1 (explanatory variable: species identity), representing species coefficients and (b) residuals from model 2 (explanatory variables: body size, top 1 value from last layer in convolutional neural network, number of species in genus, and number of training images) representing residuals for images. Species identity and number of species in genus were strongly linked, in order to keep species identity separate from other explanatory variables two models were used and residuals from model 2 compared with coefficients from model 1 compared on genus level. Higher residual values indicate that other explanatory variables than included in model 2, explain more of the variation