关于「实战工具与检查器规格:声誉与受众主权台账、预注册检验与推翻条件」,《声誉反脆弱与受众主权》给出的定量结论是什么?

结论:第八章给出蒸馏小工具「差评公开回应反转条件与交付哈希证据链检查器」、按四件套分组的检查器规格、台账字段与预注册方案:检出 0.12 星的变化每组需要 1,598 条评价。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。

数据口径:书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,不是法律意见。 书中以标注框区分已发表的证据、可运行的 Python 计算(22 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: 0.12, 1598

A. rating sd 1.210; z sum 2.8016
reviews per arm to detect 0.12 stars1,598
B. design effect 2.95; shops per arm at 40 reviews each118
C. with100 reviews per arm the smallest detectable shift is 0.480
C. with300 reviews per arm the smallest detectable shift is 0.277
C. with 1,000 reviews per arm the smallest detectable shift is 0.152
D. survey0.40 vs 0.50 willing to buy needs 388 people per arm
查看完整验算程序
# Chapter 8. p82: sample sizes for the pre-registered tests.
# Test 1: does a tailored, checkable reply policy raise later ratings
# by DELTA = 0.12 stars? Sigma is the sd of one posted rating under
# the p11 shares. Two-sided alpha 0.05, power 0.80, equal arms.
# Test 1b: the same with shops as clusters of M reviews and an
# intraclass correlation ICC (design effect 1 + (M - 1) ICC).
# Test 2: a bystander survey: share willing to buy after seeing the
# review with reply style X versus a template, 0.40 against 0.50.
# DELTA, ICC, M and the shares are illustrative planning values.
import math
from statistics import NormalDist

SHARES = (0.054, 0.095, 0.132, 0.252, 0.466)
DELTA, ALPHA, POWER = 0.12, 0.05, 0.80
M, ICC = 40, 0.05
P1, P2 = 0.40, 0.50


def z(p):
    return NormalDist().inv_cdf(p)


if __name__ == "__main__":
    tot = math.fsum(SHARES)
    p = [x / tot for x in SHARES]
    mu = math.fsum((i + 1) * x for i, x in enumerate(p))
    var = math.fsum((i + 1 - mu) ** 2 * x for i, x in enumerate(p))
    sd = math.sqrt(var)
    k = z(1.0 - ALPHA / 2.0) + z(POWER)
    n1 = math.ceil(2.0 * k * k * var / DELTA ** 2)
    print("A. rating sd {:.3f}; z sum {:.4f}".format(sd, k))
    print("   reviews per arm to detect {:.2f} stars: {:,}".format(
        DELTA, n1))
    de = 1.0 + (M - 1) * ICC
    shops = math.ceil(n1 * de / M)
    print("B. design effect {:.2f}; shops per arm at {} reviews each: {:,}"
          .format(de, M, shops))
    for n in (100, 300, 1000):
        mde = k * sd * math.sqrt(2.0 / n)
        print("C. with {:>5,} reviews per arm the smallest detectable"
              " shift is {:.3f}".format(n, mde))
    pbar = (P1 + P2) / 2.0
    num = (z(1.0 - ALPHA / 2.0) * math.sqrt(2.0 * pbar * (1.0 - pbar))
           + z(POWER) * math.sqrt(P1 * (1.0 - P1) + P2 * (1.0 - P2)))
    n2 = math.ceil(num ** 2 / (P2 - P1) ** 2)
    print("D. survey: {:.2f} vs {:.2f} willing to buy needs {:,} people"
          " per arm".format(P1, P2, n2))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《商业声誉反脆弱柔道、恶意差评证据链反制与去平台依赖私域主权母典》第 8 章「实战工具与检查器规格:声誉与受众主权台账、预注册检验与推翻条件」(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-403 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026403,
  author    = {Fan, Yafu},
  title     = {Antifragile Reputation Judo, Evidence-Backed Smear Neutralisation and Platform-Independent Audience Sovereignty},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论