Functions to implement Q-learning for estimating optimal dynamic treatment regimes from two stage sequentially randomized trials, and to perform inference via m-out-of-n bootstrap for parameters indexing the optimal regime.