FA-74652 / Experiment statistics / Member archive

Bootstrap difference interval: The upper percentile reads one position too far · case 02

Upper bounds are systematically too high.

Member previewVariant 2 · 3 implementations · 8 checks per implementation

Case contract

rng = random.Random(seed). Each of reps replicates resamples control then treatment with replacement (rng.choices, same sizes) and records mean(t) - mean(c). After sorting, the interval is diffs[floor((1 - level)/2 * reps)] to diffs[ceil((1 + level)/2 * reps) - 1]. Return both ends rounded to 6.

Why this case matters

Bootstrap intervals are the fallback for skewed metrics; reproducibility and indexing must be exact.

One recorded failure

Sample boundary fixture

This sample comes from the broken implementation of a controlled reproducer.

Boundary fixture	Actual	Expected	Outcome
treatment larger than control	`[4.833333, 7.333333]`	`[4.833333, 7.0]`	Failed

MEMBER ARCHIVE

The complete case is available to members.

This record includes three runnable implementations, regression fixtures, execution results, and source hashes.

Member access is invitation-based. Sign in with your invited account to inspect the sources.