关于「泊松稀释:路过、进店、成交三级计数的随机规律」,《快闪店选址的引力、稀释与租金》给出的定量结论是什么?

结论:模拟中,路人数的均值为 249.767,方差与均值之比为 0.983;进店数均值 14.991,理论值 15.000;成交数均值 4.503,理论值 4.500。三个计数的方差与均值之比都接近一,正是泊松分布的特征。成交者与只逛不买者的相关系数为 -0.0120,在 6000 小时的抽样误差范围内与零无异。

数据口径:书中的街道、铺位、人流、租金与成本均为示意;全部计算由书中程序生成;本书是非个人化经营建议,不构成法律或理财建议,租约条款以上架前人工复核为准。 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,并附快闪选址打分器一件。

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: -0.0120, 0.983, 14.991, 15.000, 249.767, 4.500

Poisson passers, 6000 hours
count mean var var/mean theory mean
passers249.767 245.596 0.983 250.000
entrants14.991 14.408 0.961 15.000
buyers4.503 4.374 0.971 4.500
corr(buyers, browsers)-0.0120 (theory 0)
fixed 250 passers per hour:
buyers var/mean0.985 (theory 0.982)
corr(buyers, browsers)-0.0204 (theory -0.0283)
查看完整验算程序
# Chapter 3. p31: Poisson thinning, passers -> entrants -> buyers
# Passers arrive as a Poisson process of rate LAM per hour. Each passer
# enters independently with probability P, each entrant buys with Q.
# Theory (thinning): entrants ~ Poisson(LAM P), buyers ~ Poisson(LAM P Q),
# and buyers and browsers (entrants who leave empty-handed) are independent
# Poisson counts. With a FIXED number of passers the counts are
# multinomial instead and the two are negatively correlated.
# Seeded Monte Carlo, illustrative rates.
import math
import random

rng = random.Random(20261112)
LAM, P, Q, HOURS = 250.0, 0.06, 0.30, 6000


def poisson_hour(lam):
    t, n = 0.0, 0
    while True:
        t += rng.expovariate(lam)
        if t > 1.0:
            return n
        n += 1


def stats(xs):
    mu = math.fsum(xs) / len(xs)
    var = math.fsum((x - mu) ** 2 for x in xs) / (len(xs) - 1)
    return mu, var


def corr(xs, ys):
    mx, vx = stats(xs)
    my, vy = stats(ys)
    cov = math.fsum((x - mx) * (y - my) for x, y in zip(xs, ys))
    return cov / (len(xs) - 1) / math.sqrt(vx * vy)


def run(fixed):
    pas, ent, buy, nob = [], [], [], []
    for _ in range(HOURS):
        n = int(LAM) if fixed else poisson_hour(LAM)
        e = 0
        b = 0
        for _ in range(n):
            if rng.random() < P:
                e += 1
                if rng.random() < Q:
                    b += 1
        pas.append(n)
        ent.append(e)
        buy.append(b)
        nob.append(e - b)
    return pas, ent, buy, nob


pas, ent, buy, nob = run(False)
print("Poisson passers, {} hours".format(HOURS))
print("count      mean      var   var/mean  theory mean")
for name, xs, th in (("passers", pas, LAM), ("entrants", ent, LAM * P),
                     ("buyers", buy, LAM * P * Q)):
    mu, v = stats(xs)
    print("{:<9}{:>8.3f}{:>9.3f}{:>10.3f}{:>12.3f}"
          .format(name, mu, v, v / mu, th))
print("corr(buyers, browsers) = {:+.4f}  (theory 0)".format(
    corr(buy, nob)))
pas, ent, buy, nob = run(True)
print("fixed {} passers per hour:".format(int(LAM)))
mu, v = stats(buy)
print("buyers var/mean = {:.3f}  (theory {:.3f})".format(v / mu,
      1.0 - P * Q))
p1, p2 = P * Q, P * (1.0 - Q)
th = -math.sqrt(p1 * p2 / ((1.0 - p1) * (1.0 - p2)))
print("corr(buyers, browsers) = {:+.4f}  (theory {:+.4f})".format(
    corr(buy, nob), th))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《快闪店赫夫引力选址、步行客流泊松稀释与中心地等级租金梯度母典》第 3 章「泊松稀释:路过、进店、成交三级计数的随机规律」(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-214 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026214,
  author    = {Fan, Yafu},
  title     = {Huff Gravity Siting for Pop-Up Stores, Poisson Thinning of Foot Traffic and Central-Place Rent Gradients},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论