关于「实证闭环:一人公司体能计划卡、八周可穿戴曲线与证伪方案」,《最大摄氧量、二区与五禽戏》给出的定量结论是什么?

结论:第八章给出一人公司体能计划卡、手表单次估值误差约 5.02 的八周曲线、名为「体能分级计算器」的蒸馏小工具与预注册的证伪方案。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。

数据口径:书中参数凡未注明出处者均为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何活动对任何疾病有作用,自我试验须先咨询执业医师。 第六章为五禽戏估出示意的代谢当量,一轮约 43.75 MET 分钟,并排出周剂量。

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: 5.02

absolute error SD of one watch estimate: 5.02
a reading of 35.0 means 95 % range 25.2 to 44.8
weektruewatch mean (3 readings)
135.137.0
235.438.2
335.838.3
436.137.7
536.439.0
636.739.4
737.038.3
837.339.8
fitted gain over 56 days: 2.19 (true 2.5)
readings/weekSE of 8-week gainpower (one-sided 5 %)
11.480.517
30.850.904
70.560.998
查看完整验算程序
# Chapter 8. p81: An 8-week wearable VO2max curve: what it can and cannot
# show. Molina-Garcia et al. (2022) found, for exercise-based watch
# estimates, a bias of -0.09 and 95 % limits of agreement of -9.92 to 9.74
# mL/kg/min against laboratory tests. We read the absolute error SD from
# those limits (resting-based watches: bias 2.17, limits -13.07 to 17.41,
# not used here). For change over time we assume (illustratively) that a
# person's device bias is stable and day-to-day noise has SD 1.2; the true
# value rises 2.5 over 56 days. One seeded person, 3 readings a week, then
# the analytic power of a one-sided slope test at 1, 3 and 7 readings.
import math
import random
from statistics import NormalDist

rng = random.Random(20261112)
ND = NormalDist()
SD_ABS = (9.74 - -9.92) / (2 * ND.inv_cdf(0.975))
START, GAIN, DAYS, NOISE, BIAS = 35.0, 2.5, 56, 1.2, 1.8


def days_for(k):
    return [int(round(w * 7 + j * 7 / k)) for w in range(8)
            for j in range(k)]


def slope(ts, ys):
    tb, yb = math.fsum(ts) / len(ts), math.fsum(ys) / len(ys)
    sxx = math.fsum((t - tb) ** 2 for t in ts)
    return math.fsum((t - tb) * (y - yb) for t, y in zip(ts, ys)) / sxx, sxx


if __name__ == "__main__":
    print("absolute error SD of one watch estimate: %.2f" % SD_ABS)
    print("a reading of 35.0 means 95 %% range %.1f to %.1f"
          % (35 - 1.96 * SD_ABS, 35 + 1.96 * SD_ABS))
    ts = days_for(3)
    ys = [START + BIAS + GAIN * t / DAYS + rng.gauss(0, NOISE) for t in ts]
    print("week   true   watch mean (3 readings)")
    for w in range(8):
        sel = [y for t, y in zip(ts, ys) if w * 7 <= t < w * 7 + 7]
        print("%4d   %4.1f   %4.1f" % (w + 1, START + GAIN * (w * 7 + 3)
                                       / DAYS, math.fsum(sel) / len(sel)))
    b, _ = slope(ts, ys)
    print("fitted gain over 56 days: %.2f (true %.1f)" % (b * DAYS, GAIN))
    print("readings/week   SE of 8-week gain   power (one-sided 5 %)")
    for k in (1, 3, 7):
        _, sxx = slope(days_for(k), [0.0] * (8 * k))
        se = NOISE / math.sqrt(sxx) * DAYS
        pw = ND.cdf(GAIN / se - ND.inv_cdf(0.95))
        print("%14d   %17.2f   %.3f" % (k, se, pw))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《最大摄氧量死亡率剂量响应、二区有氧乳酸阈与五禽戏间歇代谢当量母典》第 8 章「实证闭环:一人公司体能计划卡、八周可穿戴曲线与证伪方案」(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-187 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026187,
  author    = {Fan, Yafu},
  title     = {The Mortality Dose-Response of VO2max, Zone-2 Lactate Thresholds and the Metabolic Equivalents of the Five-Animal Frolics},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论