关于「每周看一次:成组序贯设计与 α 消耗函数」,《A/B 测试的偷看与小流量》给出的定量结论是什么?

结论:第五章给出成组序贯门槛,五次查看时 Pocock 为 2.4133、O'Brien-Fleming 最后一次为 2.0402。

数据口径:书中的站点、流量、转化率与提升幅度均为示意参数;全部计算由书中程序生成;本书不是个人化的经营、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(19 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: 2.0402, 2.4133

two-sided alpha 0.05 with K equally spaced looks
KPocock cp per lookOBF Cfirst-look plast-look p
11.96000.05001.96005.00e-020.0500
22.17830.02941.97755.16e-030.0480
32.28960.02202.00425.18e-040.0451
42.36140.01822.02455.15e-050.0429
52.41330.01582.04025.06e-060.0413
102.55520.01062.08674.15e-110.0369
查看完整验算程序
# Chapter 5. p51: Pocock and O'Brien-Fleming bars that keep alpha at 0.05
# Model: K equally spaced looks at accumulating normal data, two-sided.
# Pocock (1977): the same bar c at every look. O'Brien and Fleming
# (1979): bar C * sqrt(K / k) at look k, very strict early. Both
# constants are found by bisection so that the exact crossing
# probability under H0 (p11's grid recursion) equals 0.05.
import math
from statistics import NormalDist

N01 = NormalDist()
H = 0.05          # grid width in standard deviations
WIDTH = 7.0       # kernel cut at 7 sd


def cross_curve(bounds, h=H):
    """Cumulative P(|S_k| >= b_k*sqrt(k) at some look <= k)."""
    cdf = N01.cdf
    m = int(WIDTH / h) + 1
    kern = [cdf((d + .5) * h) - cdf((d - .5) * h)
            for d in range(-m, m + 1)]
    mass, n0 = [1.0], 0
    out = []
    for k, b in enumerate(bounds, 1):
        lim = b * math.sqrt(k)
        n = int(lim / h + 0.5)
        new = []
        for i in range(-n, n + 1):
            lo = max((i - .5) * h, -lim)
            hi = min((i + .5) * h, lim)
            if hi <= lo:
                new.append(0.0)
                continue
            full = lo == (i - .5) * h and hi == (i + .5) * h
            acc = []
            for jj, w in enumerate(mass):
                d = i - (jj - n0)
                if w == 0.0 or d < -m or d > m:
                    continue
                if full:
                    acc.append(w * kern[d + m])
                else:
                    u = (jj - n0) * h
                    acc.append(w * (cdf(hi - u) - cdf(lo - u)))
            new.append(math.fsum(acc))
        mass, n0 = new, n
        out.append(1.0 - math.fsum(mass))
    return out


def bisect(f, lo, hi, target=0.05, it=28):
    for _ in range(it):
        mid = (lo + hi) / 2
        if f(mid) > target:
            lo = mid
        else:
            hi = mid
    return (lo + hi) / 2


def main():
    print("two-sided alpha 0.05 with K equally spaced looks")
    print(" K   Pocock c  p per look   OBF C  first-look p  last-look p")
    for k in (1, 2, 3, 4, 5, 10):
        c = bisect(lambda x: cross_curve([x] * k)[-1], 1.5, 3.5)
        obf = bisect(lambda x: cross_curve(
            [x * math.sqrt(k / j) for j in range(1, k + 1)])[-1],
            1.5, 3.5)
        pp = 2 * (1 - N01.cdf(c))
        p1 = 2 * (1 - N01.cdf(obf * math.sqrt(k)))
        pk = 2 * (1 - N01.cdf(obf))
        print("{:2d}   {:8.4f}  {:10.4f}  {:6.4f}  {:12.2e}  {:11.4f}".format(
            k, c, pp, obf, p1, pk))


main()

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《A/B测试偷看早停假阳性膨胀、序贯概率比检验与小流量站点最小样本母典》第 5 章「每周看一次:成组序贯设计与 α 消耗函数」(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-217 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026217,
  author    = {Fan, Yafu},
  title     = {Peeking-Inflated False Positives in A/B Tests, Sequential Probability Ratio Tests and Minimum Samples for Low-Traffic Sites},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论