关于「双次抽样与平均样本量:先看少数,必要时再多看」,《验收抽样与到货验货》给出的定量结论是什么?

结论:第五章说明双次抽样在同样保护下平均只需检验 103.2 件,比单次方案少 28.8 件。

数据口径:书中批量、不良率与成本均为示意参数;全部计算由书中程序生成;本书属于非个人化经营建议。 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。

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: 103.2, 28.8

n1c1n2c2ASN(0.01)ASN(0.05)Pa(0.01)Pa(0.05)
690693103.2104.60.95200.0949
881884107.2130.90.97240.0953
7911585108.6190.80.97720.0991
5801164109.2149.10.97210.0967
8211646114.5214.00.99030.0996
single plan n = 132, c = 3; double saves 28.8 items at p = 0.01
查看完整验算程序
# Chapter 5. p51: a double plan with the same two risks, fewer items.
# Model: binomial. Draw n1; accept if d1 <= c1, reject if d1 > c2;
# otherwise draw n2 more and accept if d1 + d2 <= c2. Then
# Pa = P(d1 <= c1) + sum over c1 < k <= c2 of P(d1 = k) P(d2 <= c2-k)
# and ASN = n1 + n2 P(c1 < d1 <= c2). Search n2 = n1 or 2 n1 for the
# plan meeting Pa(0.01) >= 0.95 and Pa(0.05) <= 0.10 with the least
# ASN at p = 0.01 (best n1 kept for each c1, c2, n2 / n1).
# The single plan n = 132, c = 3 comes from p31.
import math


def pmf(n, k, p):
    return math.comb(n, k) * p ** k * (1 - p) ** (n - k)


def cdf(n, k, p):
    if k < 0:
        return 0.0
    return math.fsum(pmf(n, j, p) for j in range(min(k, n) + 1))


def double(p, n1, c1, n2, c2):
    first = cdf(n1, c1, p)
    mid = [pmf(n1, k, p) * cdf(n2, c2 - k, p)
           for k in range(c1 + 1, c2 + 1)]
    go = math.fsum(pmf(n1, k, p) for k in range(c1 + 1, c2 + 1))
    return first + math.fsum(mid), n1 + n2 * go


if __name__ == "__main__":
    best = {}
    for c1 in range(0, 4):
        for c2 in range(c1 + 1, 9):
            for k in (1, 2):
                for n1 in range(c2 + 1, 140):
                    a, asn = double(0.01, n1, c1, k * n1, c2)
                    b, _ = double(0.05, n1, c1, k * n1, c2)
                    row = (round(asn, 6), n1, c1, k * n1, c2)
                    key = (c1, c2, k)
                    if a >= 0.95 and b <= 0.10:
                        if key not in best or row < best[key]:
                            best[key] = row
    best = sorted(best.values())
    print("  n1  c1   n2  c2   ASN(0.01)  ASN(0.05)  Pa(0.01)  Pa(0.05)")
    for asn, n1, c1, n2, c2 in best[:5]:
        a, _ = double(0.01, n1, c1, n2, c2)
        b, asn5 = double(0.05, n1, c1, n2, c2)
        print("  {:3d}  {}   {:3d}   {}    {:6.1f}     {:6.1f}    {:.4f}"
              "    {:.4f}".format(n1, c1, n2, c2, asn, asn5, a, b))
    asn, n1, c1, n2, c2 = best[0]
    print("single plan n = 132, c = 3; double saves {:.1f} items at "
          "p = 0.01".format(132 - asn))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《验收抽样方案与操作特性曲线、买手到货验货抽检件数母典》第 5 章「双次抽样与平均样本量:先看少数,必要时再多看」(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-329 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026329,
  author    = {Fan, Yafu},
  title     = {Acceptance Sampling Plans and Operating-Characteristic Curves: How Many Items a Buyer Must Inspect on Arrival},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论