关于「误差注入:万分之几的年龄夸大如何把曲线压平」,《死亡率倍增与蓝区审计》给出的定量结论是什么?

结论:经过核验的意大利与国际数据显示,105 岁或 110 岁以后的死亡力大致平坦,年死亡概率约为一半。误差注入模型显示,每万条记录中极少的年龄夸大就能把一条冈珀茨曲线压成近似平台,队列混合则能制造百岁以下的减速。但制造一个 0.645 的平台需要多少错误,强烈依赖夸大幅度:小幅夸大所需的错误率高得难以置信,大幅夸大(例如身份替换)只需极少。下一章转向另一种可能:平台不是错误,而是真实的人群异质性。

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: 100, 105, 110

observed hazard by age; columns are error rates e
age0e+001e-051e-041e-03
950.3300.3300.3290.328
1000.5520.5520.5500.538
1050.9230.9200.8900.692
1101.5451.1490.5530.378
1152.5860.5630.5620.562
1204.3290.9250.9250.925
e=1e-05: log-hazard slope over 105-115 = -0.090
e=1e-04: log-hazard slope over 105-115 = -0.086
e=1e-03: log-hazard slope over 105-115 = -0.021
查看完整验算程序
# Chapter 4. p41: Rare age exaggeration bends a Gompertz curve flat.
# True lifetimes follow the Barbi et al. (2018) 65-80 line extended
# upward (m65 = 0.015, B = 0.103), with no plateau at all. A fraction
# e of records add k years to the age, k spread evenly over 5, 10 and
# 15 years. The observed hazard is f_obs(x) / S_obs(x) with
#   S_obs(x) = (1-e) S(x) + e * mean_k S(x-k),
# and f_obs the matching mixture of densities: no simulation noise.
# e values are illustrative (示意); Newman (2018, PLoS Biology) argued
# rates below 1 in 10,000 suffice to produce plateaus.
import math

M65, B = 0.015, 0.103
SHIFTS = (5, 10, 15)


def surv(x):
    if x < 65:
        return 1.0
    return math.exp(-(M65 / B) * (math.exp(B * (x - 65)) - 1.0))


def dens(x):
    if x < 65:
        return 0.0
    return M65 * math.exp(B * (x - 65)) * surv(x)


def hazard(x, e):
    s = (1 - e) * surv(x) + e * math.fsum(
        surv(x - k) for k in SHIFTS) / len(SHIFTS)
    f = (1 - e) * dens(x) + e * math.fsum(
        dens(x - k) for k in SHIFTS) / len(SHIFTS)
    return f / s


RATES = (0.0, 1e-5, 1e-4, 1e-3)
print("observed hazard by age; columns are error rates e")
print("age " + "".join("{:>10.0e}".format(e) for e in RATES))
for x in range(95, 121, 5):
    print("{:3d} ".format(x) + "".join(
        "{:10.3f}".format(hazard(x, e)) for e in RATES))
for e in RATES[1:]:
    xs = [105 + 0.5 * i for i in range(21)]
    ys = [math.log(hazard(x, e)) for x in xs]
    mx, my = math.fsum(xs) / 21, math.fsum(ys) / 21
    b = math.fsum((x - mx) * (y - my) for x, y in zip(xs, ys)) / \
        math.fsum((x - mx) ** 2 for x in xs)
    print("e={:.0e}: log-hazard slope over 105-115 = {:+.3f}".format(e, b))

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 4 章「误差注入:万分之几的年龄夸大如何把曲线压平」(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