A clinical researcher measures the reduction in blood pressure for two independent groups of patients—one receiving Drug A and the other receiving a placebo. To assess uncertainty and test for a treatment effect, she considers two resampling approaches: (i) Bootstrapping: Resample with replacement within each group to build a distribution of the difference in means; (ii) Randomization (permutation) test: Pool all observations, then randomly shuffle (without replacement) the group labels to approximate the null distribution of no treatment effect. What is the fundamental difference between these two resampling methods?
