Figure 2.

Comparative performance of RFPM and logistic regression under an interaction model. A main effects logistic regression model (LR1), a main effects + 2-way interaction logistic regression model (LR2) and a Random Forest Probability Machine (RF) are fit to 1000 simulated data sets under Model 2 (data with 3 main effects and 2 two-way interactions) under a logistic data generating model). Figure 2(a) Average conditional probability estimates of LR1, LR2 and RF are compared with the true probabilities in the data generating model. Figure 2(b) Width of the interval defined by the 5^th and 95^th percentiles of the simulation distribution of each probability target over 1000 simulations are compared between the three models that are fit.