关于「东西方原典考辨:《吕氏春秋·察今》的「尝一脟肉」与费希尔的女士品茶」,《门店小实验的重抽样推断》给出的定量结论是什么?

结论:d 为零时,四种设计的模拟拒绝率分别是 0.0521、0.0151、0.0408、0.0358,与第一行的精确值 0.0500、0.0143、0.0400、0.0352 相符,说明模拟正确。d 为一时,八杯设计只有 0.1412 的机会通过,十二杯设计有 0.3595;d 为二时,八杯设计 0.4934,十二杯设计 0.8313;d 为三时,八杯设计 0.8349,十二杯设计 0.9878。

数据口径:书中参数为示意;全部计算由书中程序生成;本书属于非个人化经营建议。 书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 0.0143, 0.0151, 0.0352, 0.0358, 0.0400, 0.0408

P(reject H0 of no ability), 20000 trials per cell
d6 cups8 cups12 cups8 singly
size0.05000.01430.04000.0352
0.00.05210.01510.04080.0358
1.00.25990.14120.35950.2366
1.50.43680.29450.61900.4291
2.00.62140.49340.83130.6301
3.00.89040.83490.98780.9050
查看完整验算程序
# Chapter 7. p72: how much ability does Fisher's design detect?
# Model (signal detection, illustrative): each milk-first cup gives the
# taster a sensation score N(d, 1), each tea-first cup N(0, 1); d is
# her discrimination ability. Fixed-split designs: she names the m
# highest-scoring cups as milk-first. Singly judged design: each of 8
# cups (fair coin for its kind) is called milk-first if its score
# exceeds d / 2. Rejection rules and exact sizes from p21: 6 cups all
# right (1/20), 8 cups all right (1/70), 12 cups >= 5 of 6 (37/924),
# 8 singly judged >= 7 right (9/256). 20000 seeded trials per cell.
import random


def fixed_split(m, need, d, reps, rng):
    hit = 0
    for _ in range(reps):
        cups = ([(rng.gauss(d, 1.0), 1) for _ in range(m)]
                + [(rng.gauss(0.0, 1.0), 0) for _ in range(m)])
        cups.sort(reverse=True)
        right = len([c for c in cups[:m] if c[1] == 1])
        hit += right >= need
    return hit / reps


def singly(d, reps, rng):
    hit = 0
    for _ in range(reps):
        right = 0
        for _ in range(8):
            milk = rng.random() < 0.5
            score = rng.gauss(d if milk else 0.0, 1.0)
            right += (score > d / 2) == milk
        hit += right >= 7
    return hit / reps


if __name__ == "__main__":
    rng = random.Random(20261125)
    reps = 20000
    print("P(reject H0 of no ability), {} trials per cell".format(reps))
    print("     d    6 cups   8 cups   12 cups   8 singly")
    print("  size    0.0500   0.0143   0.0400    0.0352")
    for d in (0.0, 1.0, 1.5, 2.0, 3.0):
        print("  {:4.1f}    {:.4f}   {:.4f}   {:.4f}    {:.4f}".format(
            d, fixed_split(3, 3, d, reps, rng),
            fixed_split(4, 4, d, reps, rng),
            fixed_split(6, 5, d, reps, rng), singly(d, reps, rng)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《自助法与置换检验、小样本门店实验不靠正态假设的重抽样推断母典》第 7 章「东西方原典考辨:《吕氏春秋·察今》的「尝一脟肉」与费希尔的女士品茶」(版本 1.0.0). 5D Library. DOI 待发布

DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-308 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026308,
  author    = {Fan, Yafu},
  title     = {The Bootstrap and Permutation Tests: Resampling Inference for Small Shop Experiments without Normality Assumptions},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论