关于「HPG 轴、睾酮与挑战假说」,《神经递质与内分泌的动力系统》给出的定量结论是什么?
结论:上面的程序已经算出:\(r = -0.061\) 只解释约 0.4% 的变异,要以 80% 的检验力检出它一次,需要约两千一百人。大多数单个研究的样本远小于此。
数据口径:健康提示:本书不是医疗建议,不作诊断,不给任何药物、激素或补充剂剂量,不声称任何练习能治疗任何疾病。 书中以标注框区分已发表的证据、可运行的 Python 计算(14 段程序,全部核对)、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: -0.061, 80
| d | P(random member of higher group is higher) | overlap of distributions |
|---|---|---|
| 0.1 | 52.8 % | 96.0 % |
| 0.2 | 55.6 % | 92.0 % |
| 0.5 | 63.8 % | 80.3 % |
| 0.8 | 71.4 % | 68.9 % |
| r = -0.061: explains 0.37 % of the variance in status-relevant behaviour | ||
| participants needed for 80 % power to detect r = 0.061: 2108 |
查看完整验算程序
# What small hormone-behaviour effects mean in practice.
import math
from statistics import NormalDist
N = NormalDist()
def from_d(d):
"""Standardised mean difference -> chance that a random member of group 1 scores higher than a
random member of group 2 (common-language effect size), and the overlap of the two distributions."""
cl = N.cdf(d / 2 ** 0.5)
overlap = 2 * N.cdf(-abs(d) / 2)
return cl, overlap
print("d P(random member of higher group is higher) overlap of distributions")
for d in (0.1, 0.2, 0.5, 0.8):
cl, ov = from_d(d)
print(f"{d:4.1f} {100 * cl:5.1f} % {100 * ov:5.1f} %")
r = -0.061 # testosterone x cortisol interaction, Dekkers et al. 2019 meta-analysis
print(f"\nr = {r}: explains {100 * r * r:.2f} % of the variance in status-relevant behaviour")
# People needed to detect such an effect once, two-sided alpha 0.05, 80 % power (Fisher z approximation)
z = 0.5 * math.log((1 + abs(r)) / (1 - abs(r)))
n = ((N.inv_cdf(0.975) + N.inv_cdf(0.80)) / z) ** 2 + 3
print(f"participants needed for 80 % power to detect r = {abs(r)}: {math.ceil(n)}")唯一数据源与引用
范雅福(Fan, Y.). (2026). 《神经递质与内分泌的动力系统》第 3 章「HPG 轴、睾酮与挑战假说」(版本 1.0.1). 5D Library. https://doi.org/10.5281/zenodo.23765340
DOI: 10.5281/zenodo.23765340 · 阅读本章所在专著 · 5DL-B-014 · v1.0.1 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026014,
author = {Fan, Yafu},
title = {Dynamical Systems of Neurotransmitters and Endocrinology},
year = {2026},
publisher = {5D Library},
edition = {1.0.1},
doi = {10.5281/zenodo.23765340},
url = {https://doi.org/10.5281/zenodo.23765340}
}