关于「实证闭环:新城市首批据点间距规划与预注册检验」,《图灵斑图与据点间距》给出的定量结论是什么?
结论:第八章给出八步规划流程、据点间距波长计算器与预注册检验:间距对半径的回归斜率检验在 15 座城市时功效 0.989。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。
数据口径:书中半径、衰减率、城市长度与数据均为示意;全部计算由书中程序生成;本书属非个人化经营方法,不构成投资、法律或选址决策建议。 书中以标注框区分已发表的证据、可运行的 Python 计算(17 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: 0.989, 15
| cities critical mean CV power at CV 0.15 | 0.30 0.50 |
| 1 | 0.436 1.000 0.928 0.554 |
| 2 | 0.554 1.000 0.999 0.850 |
| 3 | 0.607 1.000 1.000 0.954 |
| 5 | 0.666 1.000 1.000 0.997 |
| 8 | 0.703 1.000 1.000 1.000 |
| cities reject slope 0 | truth 1 reject slope 1 | truth 0.5 | |
| 6 | 0.654 0.323 |
| 10 | 0.921 0.621 |
| 15 | 0.989 0.824 |
| 25 | 1.000 0.971 |
| 40 | 1.000 0.998 |
查看完整验算程序
# Chapter 8. p82: sample size for a pre-registered test of regular
# spacing. Statistic: the gap coefficient of variation (CV) of the
# activity peaks along a corridor, averaged over m cities with 6 peaks
# each. Null: peaks fall at random (uniform positions on the corridor),
# which gives a CV near 0.85 for 5 gaps. Alternatives: regular gaps
# with log-normal scatter and CV 0.15, 0.30 or 0.50 (the seeds of p32
# gave 0.073). One-sided test at 5 per cent, 20,000 null draws and
# 4,000 draws per alternative, fixed seed. Illustrative design values.
import math
import random
rng = random.Random(20261118)
K = 6
def cv(g):
m = math.fsum(g) / len(g)
return math.sqrt(math.fsum((x - m) ** 2 for x in g) / (len(g) - 1)) / m
def null_city():
xs = sorted(rng.random() for _ in range(K))
return cv([xs[i + 1] - xs[i] for i in range(K - 1)])
def alt_city(target):
s = math.sqrt(math.log(1.0 + target * target))
return cv([math.exp(rng.gauss(0.0, s)) for _ in range(K - 1)])
print("cities critical mean CV power at CV 0.15 0.30 0.50")
for m in (1, 2, 3, 5, 8):
null = sorted(math.fsum(null_city() for _ in range(m)) / m
for _ in range(20000))
crit = null[int(0.05 * len(null))]
pw = []
for t in (0.15, 0.30, 0.50):
hits = 0
for _ in range(4000):
if math.fsum(alt_city(t) for _ in range(m)) / m < crit:
hits += 1
pw.append(hits / 4000.0)
print("%6d %16.3f %16.3f %5.3f %5.3f"
% (m, crit, pw[0], pw[1], pw[2]))
# Part 2: regularity alone does not single out Turing (central-place
# and catchment rules also space stores evenly). The sharper test is
# the scale: across m cities regress log observed spacing on log
# predicted spacing 2 pi sqrt(la_hat lh_hat). Predicted log spacings
# vary with sd 0.25 between cities, carry estimation error sd 0.06,
# and observed spacing scatters with sd 0.15 (illustrative). Report the
# chance to reject slope 0 when the truth is 1, and to reject slope 1
# when the truth is 0.5 (two-sided 5 per cent, Cornish-Fisher t).
z = 1.959963984540054
def tcrit(df):
return (z + (z ** 3 + z) / (4 * df)
+ (5 * z ** 5 + 16 * z ** 3 + 3 * z) / (96 * df * df))
def slope_test(m, beta):
xt = [rng.gauss(0.0, 0.25) for _ in range(m)]
x = [v + rng.gauss(0.0, 0.06) for v in xt]
y = [beta * v + rng.gauss(0.0, 0.15) for v in xt]
mx, my = math.fsum(x) / m, math.fsum(y) / m
sxx = math.fsum((v - mx) ** 2 for v in x)
b = math.fsum((u - mx) * (w - my) for u, w in zip(x, y)) / sxx
rss = math.fsum((w - my - b * (u - mx)) ** 2 for u, w in zip(x, y))
se = math.sqrt(rss / (m - 2) / sxx)
return b, se, tcrit(m - 2)
print("cities reject slope 0 | truth 1 reject slope 1 | truth 0.5")
for m in (6, 10, 15, 25, 40):
r0 = r1 = 0
for _ in range(4000):
b, se, tc = slope_test(m, 1.0)
r0 += abs(b) / se > tc
b, se, tc = slope_test(m, 0.5)
r1 += abs(b - 1.0) / se > tc
print("%6d %23.3f %25.3f" % (m, r0 / 4000.0, r1 / 4000.0))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《图灵反应扩散斑图、小众口碑品牌陌生城市据点涌现与区域垄断扩散比条件母典》第 8 章「实证闭环:新城市首批据点间距规划与预注册检验」(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-247 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026247,
author = {Fan, Yafu},
title = {Turing Reaction-Diffusion Patterns: How the Outposts of a Niche Word-of-Mouth Brand Emerge in a New City, and the Diffusion-Ratio Condition for Regional Dominance},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}