关于「鉴定即贝叶斯更新:四个特征的似然比、渠道基率与序贯读法」,《二手精品鉴定与柠檬折价》给出的定量结论是什么?

结论:第一章把鉴定写成贝叶斯更新,说明渠道基率的决定作用:同一条「四中二」规则,在百分之一假货的渠道里阳性预测值只有 0.223。

数据口径:书中参数为示意;全部计算由书中程序生成;本书不是对任何具体交易的定价意见,也不构成个人化的法律或理财建议。 书中以标注框区分已发表的证据、可运行的 Python 计算(20 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: 0.223

rule >2 of 4: sensitivity 0.9478, specificity 0.9666
channel share PPV NPV false flags per 1000 authentic
0.010.223 0.99945 33.4
0.030.467 0.99833 33.4
0.100.759 0.99404 33.4
0.300.924 0.97738 33.4
0.500.966 0.94876 33.4
same pattern 1100 under different channel shares:
share 0.01: P(counterfeit)0.0761
share 0.10: P(counterfeit)0.4752
share 0.50: P(counterfeit)0.8907
rule >1 of 4: sensitivity 0.9964, specificity 0.7049
rule >3 of 4: sensitivity 0.7202, specificity 0.9985
rule >4 of 4: sensitivity 0.2856, specificity 1.0000
查看完整验算程序
# Chapter 1. p12: the base rate of the channel decides what a positive
# verdict means.  Rule "flag as counterfeit when 2 or more of the four
# features fire"; its sensitivity and specificity come from exact
# enumeration of the 16 patterns under independence.  Feature values
# are the illustrative ones of p11; channel shares are illustrative.
import itertools
import math

FEATURES = [(0.80, 0.95), (0.70, 0.90), (0.60, 0.85), (0.85, 0.97)]


def rule_rates(k):
    p_flag_fake, p_flag_auth = [], []
    for pat in itertools.product((0, 1), repeat=4):
        pf = math.prod(se if b else 1 - se
                       for b, (se, sp) in zip(pat, FEATURES))
        pa = math.prod(1 - sp if b else sp
                       for b, (se, sp) in zip(pat, FEATURES))
        if pat.count(1) >= k:
            p_flag_fake.append(pf)
            p_flag_auth.append(pa)
    return math.fsum(p_flag_fake), 1.0 - math.fsum(p_flag_auth)


se, sp = rule_rates(2)
print("rule >= 2 of 4: sensitivity {:.4f}, specificity {:.4f}".format(
    se, sp))
print("channel share  PPV     NPV      false flags per 1000 authentic")
for prev in (0.01, 0.03, 0.10, 0.30, 0.50):
    ppv = se * prev / (se * prev + (1 - sp) * (1 - prev))
    npv = sp * (1 - prev) / (sp * (1 - prev) + (1 - se) * prev)
    print("   {:5.2f}      {:.3f}   {:.5f}   {:5.1f}".format(
        prev, ppv, npv, 1000 * (1 - sp)))
print("same pattern 1100 under different channel shares:")
lr = (0.80 / 0.05) * (0.70 / 0.10) * (0.40 / 0.85) * (0.15 / 0.97)
for prev in (0.01, 0.10, 0.50):
    odds = prev / (1 - prev) * lr
    print("   share {:.2f}: P(counterfeit) = {:.4f}".format(
        prev, odds / (1 + odds)))
for k in (1, 3, 4):
    a, b = rule_rates(k)
    print("rule >= {} of 4: sensitivity {:.4f}, specificity {:.4f}"
          .format(k, a, b))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《二手精品多特征贝叶斯鉴定、阿克洛夫柠檬折价均衡与限量包款二级市场折旧半衰期》第 1 章「鉴定即贝叶斯更新:四个特征的似然比、渠道基率与序贯读法」(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-171 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026171,
  author    = {Fan, Yafu},
  title     = {Multi-Feature Bayesian Authentication of Pre-Owned Luxury, Akerlof Lemon-Discount Equilibria and the Resale Half-Life of Limited Bags},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论