· Bootstrap: draw B resamples with replacement from the original sample, recompute the target statistic each time, and read confidence intervals directly from the empirical quantiles (percentile method), or calibrate them with bias correction and an acceleration factor (BCa). No normality assumption needed — ideal for skewed data or small samples.

· The sample data below is normal / near-normal, so the Bootstrap results agree with the classic σ/√n·z approximation for easy cross-checking.