干支类型特异性检验的大周期协变量分层裁决:把中世纪暖期与小冰期等50-70年及百年气候段作为分层变量后,混淆矩阵主对角线显著性是否依然稳健

Written by

in

Language

Evidence level: C

干支年型与事件类型的对角线显著性,必须先在 50-70 年/百年气候段分层后依然稳健,才配得上”类型特异性”三个字。

正文

干支检验的第一道关口是类型特异性:把历史事件按类型编码(战争、起义、瘟疫、丰年、迁都……),按年份归入六十甲子,再看”某种干支年型 ↔ 某类事件”的混淆矩阵主对角线是否显著强于随机。听上去干净,其实藏着一个巨大的混杂源——气候。

过去千年的气候段是可测的、有明确时间边界。Mann 等的全球气候代理网络重建显示,中世纪气候异常期(约 950–1250 年)多地区偏暖,小冰期最冷段集中在 1400–1700 年(PMID 19965474);Trouet 等的 947 年北大西洋涛动重建显示,中世纪暖期涛动持续偏正、转入小冰期后系统性转弱(PMID 19342585);Xing 等基于 211 条树轮序列的北半球温度重建同样给出暖期 950–1150、冷期 1450–1850 的划分,且低频变率与自然辐射强迫协变(PMID 26751947)。这些 50-70 年代际波动叠加百年冷暖段,是真实存在的”大周期协变量”。

更关键的是,气候段与人类事件之间有可测的强关联。Hsiang 等汇总 60 项最严格定量研究:温度或降雨每偏离一个标准差,人际暴力频率中位数升 4%,群体冲突升 14%(PMID 24031020);Choi 等 2024 年纳入 83 项研究的系统综述与 meta 分析给出,短期均温每升 10°C,暴力犯罪风险升 9%(PMID 39404825)。也就是说,事件类型的时间分布本身就会被气候段拉动。

于是问题来了:干支 60 年周期与 50-70 年代际气候波动、百年冷暖气段在时间轴上部分重叠。如果”某类事件集中在某些干支年”的显著性,其实是”某类事件集中在某个气候段”的影子,混淆矩阵主对角线就会系统性虚高。这时不剥气候,任何”干支类型特异性”的结论都站不住。

裁决设计并不复杂:第一步,把年份按气候段分层——暖期/冷期/过渡段三类,或对序列做百年带通滤波后取残差;第二步,在每一层内部重算干支×事件类型混淆矩阵;第三步,比较各层对角线强度是否一致。若各层都对角线显著且效应量同向,说明干支归年有独立于气候段的可测信号;若显著性只集中在某一层(比如全部落在小冰期层),”干支类型特异性”就只是气候混杂的伪装。同时要保留另一边的证据:Mach 等 2019 年组织多学科专家结构化评估,结论是气候确实影响武装冲突,但社会经济水平与国家能力等因素的影响被判定为更大、气候-冲突的机制仍不确定(PMID 31189956)——连现代数据里气候都不是唯一主角,跨千年的类型检验更没有理由不做分层。

判读时还要防三种翻车模式。其一,相位对齐伪影:干支 60 年与某些气候振荡若恰好整倍数对齐,单层显著可能只是巧合同步,须做相位随机置换看零分布是否同样能跑出这个对角线强度。其二,层内样本量塌缩:按暖/冷/过渡分三层后,事件类型在每层内的格子数可能小到检验失效,此时应合并为”冷期 vs 非冷期”两分法并报告每格最小期望频数,而不是硬撑三层。其三,编码偏差回流:事件类型的归档密度本身随朝代盛衰波动(太平年记录多、乱世记录散),而朝代更替又与气候段部分共线,所以分层前要先做记录密度归一,否则剥掉气候又混入史官笔头。分层与滤波是互为校验的两条路:分层保直观、滤波保全样本,两条路结论一致才可下裁决;只走一条就宣布稳健,等于把裁决权交还给混杂。

需要明说:本文的”干支×气候段分层”这一步尚未进入 L1 对应网络——对应网络中现有的是”六十甲子/纳音合流”总课题条目与 E-040 干支五行表(待补),边表中干支边仅涉古代蚀与地球自转(A 级),均不覆盖气候段分层裁决,因此本文证据等级封顶 C。

证据与局限

已验证: 过去千年暖期/冷期的时间边界可测、可复现(PMID 19965474、19342585、26751947);气候偏离与暴力/冲突的关联有系统综述与 meta 分析支撑(PMID 24031020、39404825)。

部分验证: 气候对冲突的影响链路:专家结构化评估判定其为真实但次于社会经济因素、机制不确定(PMID 31189956)。

待验证: 气候段分层后干支类型特异性的对角线稳健性——该分层检验本身从未被执行过,本文给出的是检验设计而非结果。

相关阅读

L1 网络状态:边表”干支”仅有 1 条边(古蚀/地转,A 级,PMID 10800374);对应网络仅有”六十甲子/纳音 | 干支合流”总课题条目与 E-040(甲乙属木,干支五行表,待补/open)。本文主题(大周期协变量分层裁决)无对应边,尚未进入 L1 对应网络。

证据等级: C
参考文献: PMID 19965474, PMID 19342585, PMID 26751947, PMID 24031020, PMID 39404825, PMID 31189956

证据等级:A=有PubMed直接支撑,B=科学类比/间接证据,C=传统文化共识/初步探索


免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

Large-Cycle Covariate Stratification in Gan-Zhi Type-Specific Testing: Is Confusion Matrix Diagonal Significance Robust After Stratifying by 50–70-Year and Century-Scale Climate…

Evidence level: C

Large-Cycle Covariate Stratification Decision for Gan-Zhi Type Specificity Testing

The diagnostic significance of the diagonal between Gan-Zhi year types and event categories must remain robust after stratification by 50–70-year or century-scale climate epochs (e.g., Medieval Warm Period, Little Ice Age) to legitimately warrant the label “type-specificity.”

Evidence Level: B | Limitation: Climate stratification introduces additional variance; significance may be attenuated compared to unstratified analysis.

Body

The first checkpoint in testing the Gan-Zhi (六十甲子, a 60-year sexagenary cycle) is type specificity: encoding historical events by category (war, rebellion, plague, harvest, capital relocation, etc.) and mapping them onto the 60-year Gan-Zhi cycle, then examining whether the main diagonal of the confusion matrix linking “a specific Gan-Zhi year type ↔ a specific event type” is significantly stronger than random. This sounds clean, yet it conceals a major confounding variable: climate.

Climate over the past millennium is measurable and bounded in time. Mann et al.’s global climate proxy network reconstruction shows that the Medieval Climate Anomaly (c. 950–1250 CE) featured widespread warming, while the Little Ice Age’s coldest phase concentrated between 1400–1700 CE (PMID 19965474); Trouet et al.’s 947 CE North Atlantic Oscillation reconstruction indicates that during the Medieval Warm Period the oscillation remained persistently positive, shifting systematically to negative after entering the Little Ice Age (PMID 19342585); Xing et al.’s Northern Hemisphere temperature reconstruction based on 211 tree-ring sequences also delineates warm periods 950–1150 CE and cold periods 1450–1850 CE, with low-frequency variability co-varying with natural radiative forcing (PMID 26751947). These 50–70 year decadal fluctuations superimposed on centennial warm/cold phases constitute real “large-cycle covariates.”

More critically, climate phases and human events exhibit measurable strong associations. Hsiang et al.’s synthesis of 60 rigorous quantitative studies found that for every standard deviation deviation in temperature or rainfall, the median frequency of interpersonal violence rises 4% and group conflict rises 14% (PMID 24031020); Choi et al.’s 2024 systematic review and meta-analysis incorporating 83 studies reported that for every 1°C rise in mean temperature, the risk of violent crime increases by 9% (PMID 39404825). Thus, the temporal distribution of event types itself is pulled by climate phases.

This raises the question: the Gan-Zhi 60-year cycle partially overlaps with 50–70 year decadal climate fluctuations and centennial cold/warm phases on the timeline. If the significance of “a certain event type concentrated in certain Gan-Zhi years” is merely a shadow of “a certain event type concentrated in a certain climate phase,” the main diagonal of the confusion matrix would be systematically inflated. Without stripping out climate, any conclusion of “Gan-Zhi type specificity” is untenable.

The adjudication design is not complex: Step 1, stratify years by climate phase—warm period/cold period/transition phase—or apply a century-bandpass filter to the sequence and retain residuals; Step 2, recompute the Gan-Zhi × event type confusion matrix within each stratum; Step 3, compare whether diagonal strength is consistent across strata. If diagonals are significant in all layers with aligned effect sizes, this indicates a measurable signal of Gan-Zhi year assignment independent of climate phases; if significance concentrates only in one layer (e.g., entirely within the Little Ice Age layer), “Gan-Zhi type specificity” is merely camouflage of climate confounding. Simultaneously, retain evidence from the other side: Mach et al.’s 2019 structured assessment by multidisciplinary experts concluded that while climate does influence armed conflict, socioeconomic level and national capacity exert greater influence, and the mechanism linking climate-conflict remains uncertain (PMID 31189956)—even in modern data, climate is not the sole protagonist, so cross-millennial type testing has no reason to omit stratification.

Interpretation must also guard against three failure modes. First, phase alignment artifact: if the Gan-Zhi 60-year cycle happens to align exactly with certain climate oscillations as integer multiples, single-layer significance may reflect coincidental synchronization; phase randomization must be performed to check whether the zero distribution can reproduce this diagonal strength. Second, intra-layer sample collapse: after stratifying into warm/cold/transition layers, the number of cells for each event type within a layer may become too small for valid testing; in such cases, merge into a binary “cold vs non-cold” scheme and report minimum expected frequencies per cell rather than forcing three layers. Third, encoding bias feedback: the density of event type archival itself fluctuates with dynastic rise and fall (peaceful years yield dense records, chaotic years sparse ones), while dynastic transitions partially collinear with climate phases; therefore, before stratification, normalize record density to avoid stripping climate only to reintroduce scribe bias. Stratification and filtering are mutually validating paths: stratification preserves intuition, filtering preserves sample integrity; conclusions must align across both paths before adjudication; relying on only one path and declaring robustness effectively returns adjudication power to confounding.

It must be stated: this paper’s “Gan-Zhi × climate phase stratification” step has not yet entered the L1 correspondence network—the existing entries in the correspondence network are the “Sixty Gan-Zhi / Na Yin convergence” general topic and the E-040 Gan-Zhi Five Phases table (pending completion); the side tables involve only ancient eclipses and Earth’s rotation (A-level), not covering climate phase stratification adjudication; therefore, this paper’s evidence level is capped at C.

Evidence and Limitations

Verified: Temporal boundaries of past warm/cold periods are measurable and reproducible (PMID 19965474, 19342585, 26751947); links between climate deviation and violence/conflict are supported by systematic reviews and meta-analyses (PMID 24031020, 39404825).

Partially Verified: The pathway of climate’s influence on conflict: expert structured assessment judges it as real but secondary to socio-economic factors, with mechanisms uncertain (PMID 31189956).

Pending Verification: Diagonal robustness of Gan-Zhi type specificity after climate stratification—this stratification test itself has never been executed; this paper provides the test design rather than results.

Related Reading

L1 Network Status: The edge list “Ganzhi” contains only one edge (Ancient Eclipse/Earth Rotation, Level A, PMID 10800374); the corresponding network includes only the “Sixty Jiazi/Nayin | Ganzhi Confluence” project entry and E-040 (Jia-Yi belongs to Wood, Ganzhi Five Phases table, pending/open). This topic (Large Cycle Covariate Stratification Decision) has no corresponding edge and has not yet entered the L1 corresponding network.

Evidence Level: C
References: PMID 19965474, PMID 19342585, PMID 26751947, PMID 24031020, PMID 39404825, PMID 31189956

Evidence Level: A = direct PubMed support, B = scientific analogy/indirect evidence, C = traditional consensus/preliminary exploration.


Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *