I would be most grateful for the opportunity to discuss the utility of the bootstrap method in managing sampling error in randomized controlled trials where the treatment allocation has been administered by simple random sampling (assuming that for practical reasons, this method is the best option). I appreciate that assuming a significance level of 5%, there is a 5% chance of obtaining a false positive outcome on comparing the same factor at baseline across the treatment and control group for multiple treatment allocations applied to the same sample using simple random sampling, as the above sampling procedure cannot completely rule out covariate imbalance. Furthermore, covariate imbalance at baseline may exist in the absence of statistical significance.

While it would seem impractical to bootstrap using multiple treatment allocations, I am interested in the utility of taking multiple bootstrap samples from the treatment and control groups for a single random treatment allocation and calculating the effect size and 95% CI. This approach does not appear to gain much coverage in the medical literature. Therefore, I would welcome opinions on whether it would help to reduce the influence of covariate imbalance, thus possibly leading to a more respectable effect size and corresponding CI than that forthcoming from using a single comparison without sampling across the treatment and control groups. I hope that I have explained my ideas sufficiently clearly. I would welcome questions if that is not the case.