Figure 5
Random forest (RF) analysis for clinical outcome at 2 weeks (A,B) and etiology (C,D). A random forest analysis was performed with the clinical outcome at 2 weeks as classifier (A,B) or the etiology as classifier (C,D). (A) Variable importance for each of the possible outcomes (poor outcome, dead, good outcome, n.a.). (B) Error rate (based on out-of-the-bag cross-validation) as a function of the number of trees generated. (C) Variable importance for each of the possible etiologies (autoimmune, bacterial, fungal, TB, unknown, viral). (D) Error rate (based on out-of-the-bag cross-validation) as a function of the number of trees generated.
