Skip to main content
. 2022 Oct 12;156:186–193. doi: 10.1016/j.jpsychires.2022.10.009

Fig. 2.

Fig. 2

Random forest to classify suicide ideation (absence/presence). Note. Variables included in random forest analysis included 20 variables significantly (p < 0.05) associated with SI from logistic regression (Step 1). Variables with positive importance values increased the accuracy of the random forest algorithm; negative values decrease accuracy. Negative values are likely due to random error.