关于「实证闭环:老板的决策日志与预注册证伪方案」,《自我损耗与饥饿法官》给出的定量结论是什么?

结论:第八章给出老板的决策日志流程、蒸馏小工具「决策排序器」与预注册方案:独自记录 120 个工作日能发现的最小差距是 0.072。书中以标注框区分已发表证据、可运行的 Python 计算(18 段程序)、原典与作者假说。

数据口径:书中参数为示意;全部计算由书中程序生成;本书不是心理或医疗建议,书中流程与工具为非个人化经营建议。 本书把两项证据放到预注册复制、元分析偏差检验与案件排序模拟面前,看它们还剩多少。

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: 0.072, 120

testgap/marginn per armdecisionsworkdays
superiority0.032,9435,886736
superiority0.051,0942,188274
superiority0.1029458874
equivalence0.051,0972,194275
equivalence0.1027555069
120 workdays -> 480 per arm; detectable gap at 80% power 0.072
查看完整验算程序
# Chapter 8. p81: How many logged decisions does a fair test need?
# Outcome: a decision judged sound at the 90-day review (yes/no).
# Superiority: early-slot rate 0.80 against a late-slot rate lower by
# 0.03, 0.05 or 0.10; two-sided alpha 0.05, power 0.80. Equivalence
# (TOST, alpha 0.05 each side, power 0.80 when the true gap is zero)
# with margins 0.05 and 0.10. Eight high-stakes decisions a day, four
# early and four late, randomised within the day. Planning values only.
# 预注册检验的样本量:一位老板要记多少条决策。
import math
from statistics import NormalDist

Z = NormalDist()
P0, PER_DAY_ARM = 0.80, 4
za2, za, zb, zb2 = (Z.inv_cdf(0.975), Z.inv_cdf(0.95),
                    Z.inv_cdf(0.80), Z.inv_cdf(0.90))


def n_super(p1, p2):
    pbar = (p1 + p2) / 2
    num = (za2 * math.sqrt(2 * pbar * (1 - pbar)) +
           zb * math.sqrt(p1 * (1 - p1) + p2 * (1 - p2))) ** 2
    return math.ceil(num / (p1 - p2) ** 2)


def n_equiv(p, margin):
    return math.ceil(2 * p * (1 - p) * (za + zb2) ** 2 / margin ** 2)


print("test           gap/margin  n per arm  decisions  workdays")
for gap in (0.03, 0.05, 0.10):
    n = n_super(P0, P0 - gap)
    print("superiority      {:.2f}       {:6,d}    {:7,d}   {:6d}".format(
        gap, n, 2 * n, math.ceil(n / PER_DAY_ARM)))
for m in (0.05, 0.10):
    n = n_equiv(P0, m)
    print("equivalence      {:.2f}       {:6,d}    {:7,d}   {:6d}".format(
        m, n, 2 * n, math.ceil(n / PER_DAY_ARM)))
days = 120
n = days * PER_DAY_ARM
se = math.sqrt(2 * P0 * (1 - P0) / n)
print("{} workdays -> {} per arm; detectable gap at 80% power {:.3f}"
      .format(days, n, (za2 + zb) * se))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《自我损耗大规模复制失败、法官饥饿效应排序伪影与决策疲劳日程母典》第 8 章「实证闭环:老板的决策日志与预注册证伪方案」(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-196 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026196,
  author    = {Fan, Yafu},
  title     = {The Failed Replication of Ego Depletion, the Ordering Artefact behind the Hungry Judges and Decision-Fatigue Scheduling},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论