# Random forest “out-of-bag” ensemble

I am using the R package RandomForestSRC for random forest applications. In the manual for the main function (rfsrc) they mention a setting called ensemble which can be set to all, out-of-bag or inbag. The manual says here only: "Always use OOB values for interfence on the training data." What does this option mean? Does it mean if I use the out-of-bag option I can actually do inference on the same data afterwards with a hypothesis generated by the random forest (e.g. wilcox test etc.)?