关于「真平台还是假平台:异质性模型、队列趋势与可辨识性」,《死亡率倍增与蓝区审计》给出的定量结论是什么?

结论:伽马脆弱度模型能产生真实的平台,但最简单的版本只能渐近地接近上限,难以复制「105 岁起即平直」的形状;平台随队列降低的观察,与「错误随时间减少」的简单预测方向相反,对纯错误解释不利。斜率检验显示,冈珀茨原样延续到 110 岁的最强说法与核验数据不符,而 0.0200 到 0.0300 的小斜率在现有样本中无法与零区分。下一章把这些工具带到蓝区本身。

数据口径:书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何产品或方法的效应,自我试验须先咨询执业医师。 书中以标注框区分已发表证据、可运行计算(22 段程序)、原典与作者假说。

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: 105, 110

frailty variance s2 = B / 0.6450.1597
age Gompertz gamma-Gompertz ratio
650.0150 0.0150 1.000
750.0420 0.0403 0.960
850.1177 0.1015 0.863
950.3297 0.2216 0.672
1050.9234 0.3834 0.415
1152.5865 0.5187 0.201
m65 0.015: hazard 105 0.383, 110 0.458, 115 0.519
m65 0.012: hazard 105 0.347, 110 0.427, 115 0.494
m65 0.009: hazard 105 0.300, 110 0.383, 115 0.458
s2 must rise by 0.0032 per birth year for a cohort trend
of -0.020 in the log plateau (Barbi et al. estimate)
s20.0489 puts 0.645 at 105: ceiling 2.105, slope 105-115 0.059
Barbi et al. slope beyond 1050.013 (SE 0.017)
查看完整验算程序
# Chapter 5. p51: A real plateau from heterogeneity (gamma frailty).
# Individual hazard z * m65 * exp(B (x-65)), z ~ gamma(mean 1, var s2).
# The population hazard is
#   mu(x) = m65 exp(B t) / (1 + s2 (m65/B) (exp(B t) - 1)),  t = x-65,
# which levels off at B / s2. With B = 0.103 (Barbi et al. 2018) we
# choose s2 so the ceiling equals their plateau 0.645, and compare it
# with the pure Gompertz line (m65 = 0.015). Lower levels m65 stand in
# for later cohorts (示意). Barbi et al. also report a cohort trend of
# -0.020 per birth year (SE 0.008) in the log plateau.
import math

B, M65, PLAT = 0.103, 0.015, 0.645
S2 = B / PLAT


def gomp(x, m=M65):
    return m * math.exp(B * (x - 65))


def frail(x, m=M65, s2=S2):
    g = math.exp(B * (x - 65))
    return m * g / (1 + s2 * (m / B) * (g - 1))


print("frailty variance s2 = B / 0.645 = {:.4f}".format(S2))
print("age   Gompertz   gamma-Gompertz   ratio")
for x in range(65, 121, 10):
    print("{:3d} {:10.4f} {:16.4f} {:7.3f}".format(
        x, gomp(x), frail(x), frail(x) / gomp(x)))
for m in (0.015, 0.012, 0.009):
    print("m65 {:.3f}: hazard 105 {:.3f}, 110 {:.3f}, 115 {:.3f}".format(
        m, frail(105, m), frail(110, m), frail(115, m)))
print("s2 must rise by {:.4f} per birth year for a cohort trend"
      .format(S2 * (math.exp(0.020) - 1)))
print("of -0.020 in the log plateau (Barbi et al. estimate)")
lo, hi = 1e-4, S2
for _ in range(80):
    mid = 0.5 * (lo + hi)
    if frail(105, s2=mid) > PLAT:
        lo = mid
    else:
        hi = mid
s2b = lo
slope = (math.log(frail(115, s2=s2b)) - math.log(frail(105, s2=s2b))) / 10
print("s2 = {:.4f} puts 0.645 at 105: ceiling {:.3f}, slope 105-115 {:.3f}"
      .format(s2b, B / s2b, slope))
print("Barbi et al. slope beyond 105: 0.013 (SE 0.017)")

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《冈珀茨死亡率倍增定律、超级百岁老人年龄误报与长寿蓝区数据审计母典》第 5 章「真平台还是假平台:异质性模型、队列趋势与可辨识性」(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-193 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026193,
  author    = {Fan, Yafu},
  title     = {Gompertz Mortality Doubling, Age Misreporting among Supercentenarians and Data Audits of Longevity Blue Zones},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论