Fig. 2From: A flexible approach for variable selection in large-scale healthcare database studies with missing covariate and outcome dataPower of each of three methods, RR-BART, BI-BART and BI-XGB, for selecting each of 10 useful predictors across 250 data replications. missForest was used for imputation. The sample size n=1000. The proportion of missingness is 40% in the outcome Y and is 60% overallBack to article page