关于「三点检验的算术:临界正确数、d′ 与最小样本量」,《冷冻高脂果肉的冰晶、香气与盲测》给出的定量结论是什么?

结论:第五章给出三点检验的精确临界值与瑟斯顿模型:感官距离 d′ 为一时正确率只有 0.4180,以百分之八十的功效检出它需要 202 位评价员。

数据口径:书中参数为示意;全部计算由书中程序生成;本书不是医疗或营养建议,不宣称任何食品的健康功效,书中的定价公式也不构成个人化的财务或经营建议。 第三章说明温度波动按阿伦尼乌斯关系放大熟化,日波动振幅八开尔文时放大 1.89 倍,并以马尔可夫链模拟三个冷链等级。

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: 0.4180, 202

targetPcfirst n 0.80safe n 0.80first n 0.90
d' 1.00.4180202220283
d' 1.50.5065535771
d' 2.00.6048232330
pd 0.200.46678794117
pd 0.300.5333404253
saw-tooth at d' 1.5 (Pc 0.5065):
n60 0.842 n61 0.870 n62 0.839 n63 0.867 n64 0.836 n65 0.864
查看完整验算程序
# Chapter 5. p53: exact power and minimum sample size of the triangle
# test, and the saw-tooth that makes "the smallest n" ambiguous.
# Test: reject when X >= x_c(n), alpha = 0.05 (as in p51 and ISO 4120).
# Power at a true Pc from the Thurstonian model (p52) or from the
# guessing model Pc = 1/3 + 2/3 pd.  "first n" is the smallest n with
# power >= target; "safe n" the smallest n from which power never drops
# below the target again (checked up to n = 700).
import math

ALPHA, NMAX = 0.05, 700


def pmf_from(n, x, p):
    """binomial probabilities for k from x up to n"""
    out = [math.comb(n, x) * p ** x * (1.0 - p) ** (n - x)]
    for k in range(x, n):
        out.append(out[-1] * (n - k) / (k + 1) * p / (1.0 - p))
    return out


CRIT = {}
for n in range(3, NMAX + 1):
    probs = pmf_from(n, 0, 1.0 / 3.0)
    acc, x = 0.0, n + 1
    while x > 0 and acc + probs[x - 1] <= ALPHA:
        acc += probs[x - 1]
        x -= 1
    CRIT[n] = x


def power(n, pc):
    return math.fsum(pmf_from(n, CRIT[n], pc))


CACHE = {}


def sizes(pc, target):
    if pc not in CACHE:
        CACHE[pc] = {n: power(n, pc) for n in range(3, NMAX + 1)}
    pw = CACHE[pc]
    first = min(n for n in pw if pw[n] >= target)
    safe = NMAX
    while safe > 3 and pw[safe - 1] >= target:
        safe -= 1
    return first, safe


CASES = (("d' 1.0", 0.4180), ("d' 1.5", 0.5065), ("d' 2.0", 0.6048),
         ("pd 0.20", 1 / 3 + 2 / 3 * 0.20), ("pd 0.30", 1 / 3 + 0.2))
print("target     Pc      first n 0.80  safe n 0.80  first n 0.90")
for name, pc in CASES:
    f8, s8 = sizes(pc, 0.80)
    f9, _ = sizes(pc, 0.90)
    print("{:8s}  {:6.4f}   {:10d}   {:10d}   {:11d}"
          .format(name, pc, f8, s8, f9))
print("saw-tooth at d' 1.5 (Pc 0.5065):")
print("   " + "  ".join("n{} {:.3f}".format(n, power(n, 0.5065))
                       for n in range(60, 66)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《高脂果肉液氮速冻奥斯特瓦尔德熟化、挥发性硫醇释放动力学与三点法感官盲测统计》第 5 章「三点检验的算术:临界正确数、d′ 与最小样本量」(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-169 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026169,
  author    = {Fan, Yafu},
  title     = {Ostwald Ripening in Liquid-Nitrogen-Frozen High-Fat Fruit Pulp, Volatile Thiol Release Kinetics and Triangle-Test Sensory Statistics},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论