Predictive performance of factors identified by the six variable selection methods to distinguish residents infected with hepatitis B virus (HBV) from the HBV-free residents.

Internal and external validation methods were used to compare the variable selection methods. Internal validation: average out-of-bag (OOB) sample prediction error based on 100 replicates was used to evaluate the performance for the six methods by using training set (80% of the total samples), followed by a testing set (20% of the total samples). External validation: average 10-fold cross-validated area under the ROC curve (AUC) was used to evaluate the performance for the six methods. Variable selection methods: stepwise, stability selection, LASSO, Bolasso, two-stage hybrid and bootstrap ranking procedures. The mean of the evaluation metric and the corresponding standard deviation (SD) based on 100 replicates are presented as the mean (SD).