关于「月度复盘,与预注册证伪」,《暗投授权与钩子归因》给出的定量结论是什么?

结论:第八章给出月度复盘流程、蒸馏小工具「白名单授权定价与钩子衰减计算器」与预注册证伪方案,检验核心命题需 13 张配有留出实验之合约。书中以标注框区分已发表证据、可运行计算(20 段程序)、原典与作者假说。

数据口径:书中参数均为示意,全部计算由书中程序生成;本书是非个人化经营分析,不构成经营或法律建议。 第六章用二叉树为跳单期权定价,合理预估 20.82 已拦住示意模型中全部跳单。

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: 13

H1 clicks per arm for lift 0.20420,322
H2 hazard ratio events views (both arms)
0.902,829 7,073
0.851,189 2,973
0.80631 1,578
H3 licences needed (rho 0.8 vs 0.3)13
Fisher z gap 0.7891
H4 creative pairs needed (0.8 vs 0.5)15
查看完整验算程序
# Chapter 8. p82: pre-registered sample sizes for the book's claims.
# H1 licence lift: exposed-only holdout, base conversion 0.02 per
#    click, lift 0.204 (the break-even lift of p62); alpha 0.05
#    two-sided, power 0.80.
# H2 hooks: two hooks compared on 3-second swipes with a log-rank
#    test; Schoenfeld (1983): events d = 4 (z_a + z_b)^2 / (ln HR)^2
#    for 1:1 allocation; views = d / share of views that swipe by 3 s
#    (0.40, close to p31).
# H3 the core claim: platform ROAS tracks holdout lift poorly. The
#    claim is refuted if the correlation across licences is 0.8 or
#    more; contracts needed to tell rho = 0.8 from rho = 0.3 by
#    Fisher z, alpha 0.05 one-sided, power 0.80.
# H4 hook pairs (Chapter 3): share of pairs in which the hook with
#    higher 3 s retention also keeps more orders; tell 0.8 from 0.5 by
#    a one-sided binomial test (normal approximation).
import math
from statistics import NormalDist

Z = NormalDist().inv_cdf

if __name__ == "__main__":
    k2 = (Z(0.975) + Z(0.80)) ** 2
    p1 = 0.02
    p2 = p1 * 1.204
    n = math.ceil(k2 * (p1 * (1 - p1) + p2 * (1 - p2)) / (p2 - p1) ** 2)
    print("H1 clicks per arm for lift 0.204: {:,}".format(n))
    print("H2 hazard ratio   events   views (both arms)")
    for hr in (0.90, 0.85, 0.80):
        d = math.ceil(4 * k2 / math.log(hr) ** 2)
        print("  {:>15.2f}   {:>6,}   {:>17,}".format(
            hr, d, math.ceil(d / 0.40)))
    k1 = (Z(0.95) + Z(0.80)) ** 2
    gap = math.atanh(0.8) - math.atanh(0.3)
    m = math.ceil(k1 / gap ** 2 + 3)
    print("H3 licences needed (rho 0.8 vs 0.3): {}".format(m))
    print("   Fisher z gap {:.4f}".format(gap))
    p0, p1 = 0.5, 0.8
    num = Z(0.95) * math.sqrt(p0 * (1 - p0)) + Z(0.80) * math.sqrt(
        p1 * (1 - p1))
    pairs = math.ceil((num / (p1 - p0)) ** 2)
    print("H4 creative pairs needed (0.8 vs 0.5): {}".format(pairs))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《暗投白名单授权经济学、三秒钩子生存分析与单层星探归因博弈母典》第 8 章「月度复盘,与预注册证伪」(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-333 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026333,
  author    = {Fan, Yafu},
  title     = {Dark-Ads Whitelist Licensing Economics, Three-Second Hook Survival Analysis and Single-Tier Scout Attribution Games},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论