关于「约三倍的层比:5、15、50、150 与比例的稳健性」,《邓巴分层与私域半衰期》给出的定量结论是什么?
结论:第二章由六个平均群体规模重算层比,得到 3.258 到 3.379,并说明共同乘数因子为何在相除时被约掉。
数据口径:书中的层级下限、半衰期、权重与停止概率均为示意参数;全部计算由书中程序生成;本书是非个人化经营建议,合规规则以各地现行法规为准。 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。
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: 3.258, 3.379
| level mean size ratio to previous | |
| clique | 4.6 |
| sympathy | 14.3 3.109 |
| band | 42.6 2.979 |
| community | 132.5 3.110 |
| megaband | 566.6 4.276 |
| tribe | 1728.0 3.050 |
| mean of successive ratios | 3.305 |
| geometric mean ratio | 3.273 |
| log-linear slope ratio | 3.303 |
| fitted sizes | 4.3 14.1 46.5 153.5 507.1 1674.7 |
| leave-one-out slope ratio | 3.258 to 3.379 |
| inner four levels only | 3.057 |
| first four, ratio 3 from 5 | 5 15 45 135 |
查看完整验算程序
# Chapter 2. p21: the layer ratio re-estimated from Zhou et al. (2005)
# Data: the six mean group sizes reported by Zhou, Sornette, Hill and
# Dunbar (2005): support clique, sympathy group, band, community,
# megaband, large tribe. Three estimators of the scaling ratio: mean
# of successive ratios, geometric mean (end points only) and the slope
# of ln(size) on level by least squares; a leave-one-level-out range
# shows how much any single level moves the slope estimate.
import math
NAMES = ["clique", "sympathy", "band", "community", "megaband",
"tribe"]
MEANS = [4.6, 14.3, 42.6, 132.5, 566.6, 1728.0]
def slope_ratio(levels, sizes):
n = len(levels)
ys = [math.log(s) for s in sizes]
mx = math.fsum(levels) / n
my = math.fsum(ys) / n
sxx = math.fsum([(x - mx) ** 2 for x in levels])
sxy = math.fsum([(x - mx) * (y - my) for x, y in zip(levels, ys)])
b = sxy / sxx
a = my - b * mx
return math.exp(b), math.exp(a)
def main():
print("level mean size ratio to previous")
ratios = []
for i, (nm, m) in enumerate(zip(NAMES, MEANS)):
if i == 0:
print("{:10s} {:9.1f}".format(nm, m))
continue
r = m / MEANS[i - 1]
ratios.append(r)
print("{:10s} {:9.1f} {:6.3f}".format(nm, m, r))
print("mean of successive ratios : {:.3f}".format(
math.fsum(ratios) / len(ratios)))
geo = (MEANS[-1] / MEANS[0]) ** (1.0 / (len(MEANS) - 1))
print("geometric mean ratio : {:.3f}".format(geo))
lv = list(range(len(MEANS)))
r_all, base = slope_ratio(lv, MEANS)
print("log-linear slope ratio : {:.3f}".format(r_all))
print("fitted sizes : " + " ".join(
"{:.1f}".format(base * r_all ** k) for k in lv))
loo = []
for j in lv:
keep = [k for k in lv if k != j]
loo.append(slope_ratio(keep, [MEANS[k] for k in keep])[0])
print("leave-one-out slope ratio : {:.3f} to {:.3f}".format(
min(loo), max(loo)))
r4, b4 = slope_ratio(lv[:4], MEANS[:4])
print("inner four levels only : {:.3f}".format(r4))
print("first four, ratio 3 from 5 : 5 15 45 135")
main()唯一数据源与引用
范雅福(Fan, Y.). (2026). 《邓巴数分层、关系维护时间预算与私域客户衰减半衰期母典》第 2 章「约三倍的层比:5、15、50、150 与比例的稳健性」(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-231 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026231,
author = {Fan, Yafu},
title = {Dunbar's Layered Numbers, Relationship-Maintenance Time Budgets and the Decay Half-Life of Private-Channel Customers},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}