关于「实证闭环:前两小时估分支比,再决定是否追加推流」,《评论裂变分支比》给出的定量结论是什么?

结论:第八章给出前两小时估分支比再决定推流的闭环、小工具「裂变分支比估计器」与预注册推流实验(每组 43 条帖子,协变量调整后 13 条)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。

数据口径:书中的帖子、账号、流量、成本与评论价值均为示意参数;全部计算由书中程序生成;本书是非个人化经营建议,只讨论真实用户的评论与平台官方工具买来的曝光,不涉及任何制造评论或虚构稀缺的手段。 第四章用小崎递推、补偿子恒等式与黄金分割求最大似然估计,示意帖的分支比 95 % 区间从 0.42 到 0.92,两小时估计的标准差约 0.186。

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, 43

z_a + z_b = 2.8016
nsd ctrlpredicted liftposts/armwith R2=0.7
0.387.7128.683
0.6156.1225.083
0.8324.0450.093
to tell n = 0.6 from n = 0.3 (lift gap 96.4): 43 posts/arm, 13 adjusted
refuted if the 95% interval of lift / D excludes 1/(1-n_pooled)
查看完整验算程序
# Chapter 8. p82: Posts needed for a pre-registered push experiment.
# Design: posts randomised 1:1 to a push of D = 90 extra outside
# commenters or to none. Outcome: comments in 12 hours. Model: outside
# commenters M per post are Poisson with a mean that varies across posts
# (base mean 120, coefficient of variation 0.5); each brings a Borel
# cluster with E K = 1/(1-n), Var K = n/(1-n)^3 (p11). Compound variance
#   Var = E[M] Var K + Var[M] (E K)^2.
# Prediction to test: the push adds D/(1-n) comments, n from pooled
# pre-period fits. Two-sided alpha 0.05, power 0.8; an adjustment for
# the pre-push two-hour count with R^2 = 0.7 is shown as well.
# All inputs illustrative.
import math
from statistics import NormalDist

D, BASE, CV = 90.0, 120.0, 0.5
Z = NormalDist().inv_cdf(0.975) + NormalDist().inv_cdf(0.8)


def arm_var(m_extra, n):
    ek, vk = 1 / (1 - n), n / (1 - n) ** 3
    em = BASE + m_extra
    vm = em + (CV * BASE) ** 2
    return em * vk + vm * ek ** 2


def per_arm(delta, n, r2=0.0):
    v = (arm_var(0.0, n) + arm_var(D, n)) * (1 - r2)
    return math.ceil(Z ** 2 * v / delta ** 2)


print("z_a + z_b = %.4f" % Z)
print("n     sd ctrl  predicted lift  posts/arm  with R2=0.7")
for n in (0.3, 0.6, 0.8):
    lift = D / (1 - n)
    print("%.1f   %6.1f   %8.1f        %5d      %5d"
          % (n, math.sqrt(arm_var(0.0, n)), lift, per_arm(lift, n),
             per_arm(lift, n, 0.7)))
gap = D / (1 - 0.6) - D / (1 - 0.3)
print("to tell n = 0.6 from n = 0.3 (lift gap %.1f): %d posts/arm, %d"
      " adjusted" % (gap, per_arm(gap, 0.6), per_arm(gap, 0.6, 0.7)))
print("refuted if the 95% interval of lift / D excludes 1/(1-n_pooled)")

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《霍克斯自激点过程、评论裂变分支比临界值与直播跟风抢购余震半衰期母典》第 8 章「实证闭环:前两小时估分支比,再决定是否追加推流」(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-234 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026234,
  author    = {Fan, Yafu},
  title     = {Hawkes Self-Exciting Point Processes: Branching-Ratio Criticality in Comment Cascades and the Aftershock Half-Life of Livestream Herd Buying},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论