5D Library › 玛雅卓尔金双模齿轮、神谕十字对合变换与六十四闸门网络渗流 · English
第八章 基率、标签的信息量与两项预注册研究
巧合的基率
推导
# Chapter 8. p15: Base rates and information in the labels.
# (1) Coincidences: in a group of n, the chance that two people share a
# kin (260 equally likely), a colour (4), a type (simulated shares).
# (2) Information: the entropy in bits of each label - kin, colour, type,
# profile - i.e. how finely a label can sort people at all.
import importlib.util
import random
from math import log2, prod
def load(name):
spec = importlib.util.spec_from_file_location(name, name + ".py")
m = importlib.util.module_from_spec(spec)
spec.loader.exec_module(m)
return m
pop = load("p11_population_model")
rules = load("p09_centres_types")
def share_uniform(n, k):
return 1 - prod((k - i) / k for i in range(n)) if n <= k else 1.0
def entropy(ps):
return -sum(p * log2(p) for p in ps if p > 0)
if __name__ == "__main__":
rng = random.Random(72015)
N, kinds = 10000, {}
for _ in range(N):
k = rules.analyse(set(pop.chart(rng.uniform(2433282.5,
2462502.5))))[3]
kinds[k] = kinds.get(k, 0) + 1
tshare = [v / N for v in kinds.values()]
same_type = sum(p * p for p in tshare)
print(" two strangers share: kin {:.4f}, colour {:.4f}, type {:.4f}"
.format(1 / 260, 1 / 4, same_type))
for n in (5, 10, 20, 30):
print(" group of {:2d}: some pair shares a kin {:.3f}".format(
n, share_uniform(n, 260)))
n50 = next(n for n in range(2, 100) if share_uniform(n, 260) >= 0.5)
print(" smallest group with an even chance of a shared kin: {}".format(
n50))
prof = [0.8667 / 6] * 6 + [0.1333 / 6] * 6
print(" entropy, bits: kin {:.2f}, colour {:.2f}, type {:.2f},"
" profile {:.2f}".format(log2(260), log2(4), entropy(tshare),
entropy(prof)))
# two strangers share: kin 0.0038, colour 0.2500, type 0.2855
# group of 5: some pair shares a kin 0.038
# group of 10: some pair shares a kin 0.161
# group of 20: some pair shares a kin 0.528
# group of 30: some pair shares a kin 0.825
# smallest group with an even chance of a shared kin: 20
# entropy, bits: kin 8.02, colour 2.00, type 1.92, profile 3.15两个陌生人的 kin 相同的概率是 0.0038,颜色相同的概率是 0.2500,人类图类型相同的概率是 0.2855(按第六章的类型比例计算)。在一个 20 人的团体里,有两个人 kin 相同的概率已经过半(0.528);30 人时为 0.825。「我们竟然是同一个印记」「我们都是生产者」这类发现,在稍大的团体中几乎必然出现。
标签能区分多少
一个标签能把人分得多细,由它的熵决定。kin 有 8.02 比特,颜色 2.00 比特,人类图类型 1.92 比特,人生角色 3.15 比特(因为 12 种角色的概率不相等,小于 )。类型这样一个不到 2 比特的标签,只能把人分成大约四类;每一类里都有约四分之一以上的人。对一个覆盖四分之一人口的描述,第 71 本第六章的覆盖模型说明它几乎会被每个人接受:这是巴纳姆效应在类型描述上的形式 (Forer 1949; Dickson and Kelly 1985)。
注
四步核查。 第一,写明用的是哪一种计数(传统 GMT 584283 还是 Dreamspell,见 p03)与哪一套星历、哪一个设计时刻的算法。第二,核对出生时间的精度:一小时的误差会改变月亮的爻线(第六章)。第三,先算基率:任何「巧合」或「稀有类型」,先与本章和第五、六章的基率比较。第四,把「像不像我」换成盲法辨认。
研究甲:盲法辨认
每名参与者收到四份去掉标签的「类型与人生角色」描述,一份由自己的盘写出,三份由陌生人的盘写出(事先选定与参与者类型不同的盘),从中选出最像自己的一份。机会命中率为 1/4。
推导
# Chapter 8. p16: Two pre-registered studies and their sizes.
# (A) Blind recognition: each participant reads four type-and-profile
# descriptions written from four different charts (theirs and three
# strangers' matched to differ in type), with labels removed, and picks
# the one most like them. Chance 1/4; exact one-sided binomial test,
# alpha 0.05, power 0.90.
# (B) Oracle ties and couples: the claim that a partner who is one's
# analog, antipode, occult or guide makes a different relationship.
# One is in the other's oracle for 0.01455 of ordered pairs (p05);
# here we count unordered couples tied in either direction, and how
# many couples must be recruited to observe 150 tied couples.
from itertools import product
from math import comb, ceil
KINS = list(product(range(20), range(1, 14)))
def oracle(k):
s, t = k
return {((s + 12 * (t - 1)) % 20, t), ((19 - s) % 20, t),
((s + 10) % 20, t), ((21 - s) % 20, 14 - t)} - {k}
def tail(n, k, p):
return sum(comb(n, i) * p ** i * (1 - p) ** (n - i)
for i in range(k, n + 1))
def plan(p0, p1, alpha=0.05, power=0.90):
for n in range(20, 5000):
k = next(k for k in range(n + 1) if tail(n, k, p0) <= alpha)
if tail(n, k, p1) >= power:
return n, k, tail(n, k, p0)
if __name__ == "__main__":
for p1 in (0.32, 0.35):
n, k, a = plan(0.25, p1)
print(" (A) hit rate {:.2f} vs 0.25: {} participants, at least {}"
" hits".format(p1, n, k))
print(" exact size {:.4f}".format(a))
tied = sum(1 for a, b in product(KINS, KINS)
if a != b and (b in oracle(a) or a in oracle(b)))
either = tied / (260 * 259)
print(" (B) couple tied either way: {:.5f}".format(either))
print(" couples to recruit for 150 tied couples: {}".format(
ceil(150 / either)))
# (A) hit rate 0.32 vs 0.25: 359 participants, at least 104 hits
# exact size 0.0485
# (A) hit rate 0.35 vs 0.25: 179 participants, at least 55 hits
# exact size 0.0485
# (B) couple tied either way: 0.01752
# couples to recruit for 150 tied couples: 8561若真实命中率为 0.32,单侧 0.05、把握度 0.90 的精确二项检验需要 359 名参与者,至少 104 人命中,实际第一类错误为 0.0485;若真实命中率为 0.35,需要 179 人、至少 55 人命中。描述应由不知道参与者身份的写作者依盘写出,并删去对任何人都适用的笼统句子。
研究乙:神谕关系与伴侣
第三章算出,一对伴侣在任一方向上存在神谕关系(支持、挑战、隐藏推动或指引)的概率为 0.01752。要得到 150 对有神谕关系的伴侣,需要招募约 8561 对。结局指标(例如关系满意度量表)与分析计划必须在知道任何一对的 kin 之前登记,比较有关系与无关系的伴侣。
假说 · 尚未验证
预注册的预测(尚未验证): 研究甲的命中率与 0.25 无显著差异;研究乙中有神谕关系与无神谕关系的伴侣在预先指定的结局上无显著差异。什么会证明它错: 研究甲命中率显著高于 0.25,或研究乙中两组在预先指定的结局上有显著差异。
全书小结
卓尔金历是两个互素齿轮的乘积,历法轮与年承载者由同余算术决定;Dreamspell 是另一种计数,与传统计数从不重合。五大神谕是 上的群作用,在颜色上是克莱因四元群,指引是仿射映射。人类图的闸门轮是邵雍的先天圆图;在真实天空中反映者约占 0.0084(p11),十二种人生角色的不均匀概率来自 88 度这一个常数。这些结论都来自系统本身的结构,不涉及任何印记或类型是否应验。
参考文献
Fan, Y. (2026). Mayan Tzolkin Modular Gears, Oracle-Cross Involutions and Sixty-Four-Gate Network Percolation: Coprime Cycle Synchronization, Antipode-Analog Permutations and Nine-Center Connected Components [玛雅卓尔金双模齿轮、神谕十字对合变换与六十四闸门网络渗流] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090193 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).