Figure 5
From: A simple model of COVID-19 explains disease severity and the effect of treatments
Regression tree analysis results. (a) Single optimal tree for the prediction of hyperinflammation from model parameters. The tree reads from left to right. At each labeled node, simulations either go up if the value for the associated parameter is higher than a threshold determined by the procedure (threshold not shown, see Supplementary Figure S7), or down otherwise. Branch length represents the amount of classification error explained by the node. At each terminal node, the percentage of simulations as well as the risk of hyperinflammation within members of the node are reported. (b) Parameter importance based on the GINI index. A greater mean GINI decrease indicates a parameter that is more discriminatory.
