Quantifying effects in two-sample environmental experiments using bootstrap confidence intervals