关于「条件期望是一条斜率小于一的直线:头部门店为什么来年下滑」,《均值回归与封面诅咒》给出的定量结论是什么?
结论:第一章说明,在年间相关为 0.5 时,前十分之一门店中有 0.8380 来年成绩低于今年,没有任何人松懈;并复现塞克里斯特「平庸的胜利」与霍特林的反驳。
数据口径:书中门店、达人、顾客与报告的参数均为示意;全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营、法律或理财建议,涉及运程报告与原典之处只分析结构与概率,不作任何个人占断。 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。
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: 0.5, 0.8380
| top-decile threshold z = 1.2816 | |||||
| rho | E[X|top] | E[Y|top] | kept | P(Y<X) | P(stay top) |
| 0.2 | 1.7550 | 0.3510 | 0.20 | 0.9140 | 0.1720 |
| 0.4 | 1.7550 | 0.7020 | 0.40 | 0.8667 | 0.2665 |
| 0.5 | 1.7550 | 0.8775 | 0.50 | 0.8380 | 0.3240 |
| 0.6 | 1.7550 | 1.0530 | 0.60 | 0.8049 | 0.3902 |
| 0.8 | 1.7550 | 1.4040 | 0.80 | 0.7188 | 0.5624 |
| 0.9 | 1.7550 | 1.5795 | 0.90 | 0.6557 | 0.6886 |
| kept = E[Y|top] / E[X|top] = rho exactly; no effort term |
核心公式
\[E[Y \mid X = x] = \rho\,x .\]
查看完整验算程序
# Chapter 1. p11: What a top-decile store should expect next year.
# Model: standardised scores X (this year) and Y (next year) are
# bivariate normal with correlation rho, so E[Y | X = x] = rho * x.
# Nothing about effort changes between years. For stores in the top
# 10 % of X we compute, by exact formulas and numerical integration,
# the mean of X, the expected mean of Y, the share whose score falls,
# and the share that stays in the top 10 %. rho values are illustrative.
# 模型:双变量正态,E[Y|X]=rho*X;头部 10 % 门店来年的预期位置。
import math
def phi(z):
return math.exp(-0.5 * z * z) / math.sqrt(2.0 * math.pi)
def cdf(z):
return 0.5 * (1.0 + math.erf(z / math.sqrt(2.0)))
def quantile(p):
lo, hi = -10.0, 10.0
for _ in range(200):
mid = 0.5 * (lo + hi)
if cdf(mid) < p:
lo = mid
else:
hi = mid
return 0.5 * (lo + hi)
TOP = 0.10
C = quantile(1.0 - TOP) # threshold of the top decile
def integrate(f, a, b, n=20000):
h = (b - a) / n
return h * math.fsum(f(a + (i + 0.5) * h) for i in range(n))
def table_row(rho):
s = math.sqrt(1.0 - rho * rho)
mean_x = phi(C) / TOP
mean_y = rho * mean_x
worse = integrate(lambda x: phi(x) * cdf(x * (1 - rho) / s),
C, 9.0) / TOP
stay = integrate(lambda x: phi(x) * (1.0 - cdf((C - rho * x) / s)),
C, 9.0) / TOP
return mean_x, mean_y, worse, stay
if __name__ == "__main__":
print("top-decile threshold z = {:.4f}".format(C))
print(" rho E[X|top] E[Y|top] kept P(Y<X) P(stay top)")
for rho in (0.2, 0.4, 0.5, 0.6, 0.8, 0.9):
mx, my, w, st = table_row(rho)
print(" {:.1f} {:.4f} {:.4f} {:.2f} {:.4f} {:.4f}"
.format(rho, mx, my, my / mx, w, st))
print("kept = E[Y|top] / E[X|top] = rho exactly; no effort term")唯一数据源与引用
范雅福(Fan, Y.). (2026). 《均值回归与封面诅咒、运程报告后验校准与塞翁失马序列独立性母典》第 1 章「条件期望是一条斜率小于一的直线:头部门店为什么来年下滑」(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-202 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026202,
author = {Fan, Yafu},
title = {Regression to the Mean and the Cover Jinx: Posterior Calibration of Fortune Reports and Serial Independence in the Tale of the Lost Horse},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}