Skip to main content
. 2021 Dec 30;119(1):e2020956119. doi: 10.1073/pnas.2020956119

Fig. 3.

Fig. 3.

Dependence of model dynamics on different parameters and CFT. A shows the variable importance score for all the parameters from an RF classification (with 10,000 trees) on the three categories of spatial disturbance (OOB = 10.2%). B and C denote the variable importance scores from RF regression models (with 10,000 trees) for SD mean and variance, respectively. D, E, and F show the SD mean–variance plots overlaid with mutation size, growth radius, and CFT values of the respective simulations. For A–C, IncNodePurity (used in regression RFs) and MeanDecreaseGini (used in classification RFs) refer to how well a parameter explains the prediction variable in question.