关于「嗓音稳定度:自相关基频、抖动与语速变异系数」,《三秒薄切片信任》给出的定量结论是什么?

结论:第五章用自相关估计基频,说明抖动三秒可测、语速变异要更长录音,低音高效应在语境间差异较大时,下一个语境出现负效应的概率可达 0.120。

数据口径:书中参数为示意;全部计算由书中程序生成;本书只写如何让真实、可兑现的信号在前三秒清晰一致,不写任何伪造资质或假权威的手法,不宣称任何气味的健康功效,是非个人化经营建议。 第一章区分信度与效度,示意参数下三秒判断只达到效度上限的 0.612,与长期评价的一致率为 0.579。

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: 0.120

A. mu tau95% prediction interval P(d < 0) next context
0.300.10 [+0.08, +0.52] 0.004
0.300.25 [-0.20, +0.80] 0.120
0.200.25 [-0.30, +0.70] 0.216
0.100.25 [-0.40, +0.60] 0.347
0.000.25 [-0.50, +0.50] 0.500
B. d share choosing the lower version
0.00.500
0.10.528
0.20.556
0.30.584
0.50.638
C. listeners needed to detect the share in B (two-sided 0.05,
power 0.80, one choice per listener)
d0.1: share 0.528, n = 2,468
d0.2: share 0.556, n = 619
d0.3: share 0.584, n = 276
查看完整验算程序
# Chapter 5. p53: a lower voice as a weak, context-bound prior.
# Random-effects view: in context c the standardised effect of a lower
# pitch on perceived leadership is d_c ~ N(mu, tau^2). A pooled mean mu
# says little about the next context when tau is comparable to mu; the
# 95 % prediction interval is mu +- 1.96 sqrt(tau^2 + se^2). Part B
# converts a d into the share of listeners who prefer the lower version in
# a two-voice choice, Phi(d / sqrt(2)). Part C gives the number of
# listeners a test of that share needs. Every mu, tau and se here is
# illustrative, not a pooled estimate of any study.
import math


def phi(x):
    return 0.5 * (1.0 + math.erf(x / math.sqrt(2.0)))


if __name__ == "__main__":
    se = 0.05
    print("A.  mu    tau   95% prediction interval   P(d < 0) next context")
    for mu, tau in ((0.30, 0.10), (0.30, 0.25), (0.20, 0.25),
                    (0.10, 0.25), (0.00, 0.25)):
        s = math.sqrt(tau * tau + se * se)
        print("   {:.2f}  {:.2f}   [{:+.2f}, {:+.2f}]              {:.3f}"
              .format(mu, tau, mu - 1.96 * s, mu + 1.96 * s,
                      phi(-mu / s)))
    print("B. d     share choosing the lower version")
    for d in (0.0, 0.1, 0.2, 0.3, 0.5):
        print("   {:.1f}   {:.3f}".format(d, phi(d / math.sqrt(2.0))))
    print("C. listeners needed to detect the share in B (two-sided 0.05,")
    print("   power 0.80, one choice per listener)")
    for d in (0.1, 0.2, 0.3):
        p = phi(d / math.sqrt(2.0))
        n = ((1.959964 * 0.5 + 0.841621 * math.sqrt(p * (1.0 - p)))
             / (p - 0.5)) ** 2
        print("   d = {:.1f}: share {:.3f}, n = {:,}".format(
            d, p, math.ceil(n)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《三秒薄切片信任捕获:视听嗅三通道同频一致性与瞬间权威建立母典》第 5 章「嗓音稳定度:自相关基频、抖动与语速变异系数」(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-419 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026419,
  author    = {Fan, Yafu},
  title     = {Three-Second Thin-Slicing Trust: Visual-Auditory-Olfactory Congruence and Instant Authority Signalling},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论