Blog

  • 判读者内重测与判读者间相关的联合分解:隔周重读能把「判读者内漂移」从 ρ 里拆出多少

    Evidence level: B

    判读者之间的相关系数偏低,可能是「几个人尺子不一样」,也可能是「同一个人自己的尺子在漂」——不加重测轮,这两种原因永远混在一个数里分不开。

    正文

    前一篇判读者间相关的实测分布研究给多中心设计定下了规格:多位判读者独立判读同批样本,用判读者间相关 ρ 来量化「分歧有多大」。但那个设计有一个天然的盲区——它只分解了判读者之间的分量。如果 ρ 偏低,原因可能有两种:第一种,几位判读者的尺子本来就不一样,你严我宽,彼此错位;第二种,每一位判读者自己的尺子在时间上就不稳,今天判的和两周后判的,同一个样本给出不同的读数。这两种原因在只测一轮的设计里完全混叠,看到的低 ρ 既可能是「人不齐」,也可能是「人自己在漂」。

    把两者拆开的办法在测量学里是现成的:让至少三位判读者对同批至少三十个样本隔两周重读一遍,给每位判读者建立「判读者内重测一致性」(ICC intra),与原有的判读者间相关(ρ inter)并列比较。数学上这走的是概括理论的多层方差分解路线:把总方差拆成样本间、判读者间、判读者内时间漂移三层,各层贡献用混合模型同时估计(PMID 33818118)。这正是前一篇设计缺的那一层——加一个「时间」维度,混合的两种原因就有了各自的出口。

    这个拆分值得做吗?值得。因为「同一个人隔段时间判同一个东西会漂」在生理测量里是实测过的常态:健康成人心率变异性指标一周内重测,时间域指标一致性尚可、频域指标明显更差,静息态信度系数只在 0.50 到 0.83 之间(PMID 39854840)。也就是说,即便是训练有素的标准化测量,「同一把尺子隔一周不复原」并不是罕见事故,而是频域类读数的默认属性。判读构型这种主观成分更重的任务,没有理由假设自己比仪器读数更稳。而重测研究还有一个常被忽略的前置问题:重测信度的估计本身对样本量极其敏感,样本不足时重测系数的估计会系统性失真(PMID 40119099)——所以重测轮不是「加两个时点」那么简单,30 例 3 人的规格是按估计精度反推的下限,不是拍脑袋。

    还有一个更隐蔽的坑必须点名:普通的信度系数对「整体往上或往下漂」这类系统性偏置几乎不敏感。方法学文献早已指出,在存在系统偏置时,相关系数、配对检验与组内相关系数会给出互相矛盾的结论,而绝对一致性定义的组内相关系数是三者中唯一能反映偏置的,但仍须配均值检验确认(PMID 11091941);组内相关系数的绝对一致版与一致性版近乎相等时,才能宣布「系统偏置可忽略」(PMID 31329615)。落到本文的设计上就是一条硬规矩:判读者内重测一致性必须报绝对一致口径,只报一致性口径的重测轮等于白测——漂移恰好整体同向时它照样给你高分。

    裁决标准三分支,事先写死:若判读者内重测一致性与判读者间相关之差小于 0.1,说明时间漂移可忽略,前一篇的多中心设计不必加重测轮,省一轮成本;若两者之差在 0.1 到 0.2 之间,判读内漂移与判读者间分歧同量级,任何 ρ 的报告必须同时报两个分量、不得只报合并值;若差达到 0.2 以上,则时间漂移是主导项——任何不加重测校正的 ρ 估计都作废,单人判读的长期可用性存疑,判读规程必须改成「定期重校准」模式。

    这一层分解对易学研究为什么关键?因为整个「五脏辨证标签能不能当分组变量」的前提问题,正卡在标签信度上:开放人群里中医医师间诊断一致性被报告为低、作者称为令人担忧(PMID 30383398),但另一侧证据说训练可以改善、量表化可行。冲突双方都成立的可能性完全存在——因为「低一致性」这个观察本身可能就有一部分来自判读者内的时间漂移,而非判读者之间真实的理论分歧。不拆开,这场关于「辨证标签可不可靠」的争论就会永远在两个不同误差源的混淆里打转。

    证据与局限

    已验证:多层方差分解可同时估计判读者间与判读者内分量(概括理论框架成熟);频域生理读数一周重测信度 0.50 至 0.83、差于时间域,为「时间漂移量级真实存在」提供直接实测锚点;绝对一致性口径的组内相关系数是察觉系统偏置的必要口径。

    部分验证:重测信度估计的样本量敏感性有直接方法学文献,但「3 人 30 例」的具体规格是按本项目判读构型任务反推的,未在气功外气构型判读上实测过。

    待验证:三分支判据(差小于 0.1 免重测轮、达到 0.2 强制校正)为项目内部仿真推演,需按本设计采集真实数据后回填;判读者内漂移与判读者间分歧在真实辨证任务中的实际占比未知。

    相关阅读

    证据等级: B
    参考文献: PMID 40119099, PMID 33818118, PMID 39854840, PMID 30383398, PMID 31329615, PMID 11091941

    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

    L1 网络状态:本主题对应的「判读者内重测 × 判读者间相关联合分解」边尚未进入 L1 对应网络,已登记 science-yixue-l1-gaps.md。最近邻边为 E-060 标签前提(五脏辨证作分组标签 → 标签信度,状态 冲突(未裁决),原文摘录:「中医医师间诊断一致性在开放人群中『低(low)』,作者称为『令人担忧』(PMID 30383398);同系列第二篇指出偶然校正一致性统计常被误用、建议 Gwet AC2(PMID 31730403 / 31634000 / 31730404);但训练可改善(PMID 19538100)、量表化可行」)——本文已按规则把冲突双方(标签一致性低 vs 训练可改善)并列写出,未只挑一边。


    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。

    Joint decomposition of within-reader and between-reader correlation: how much does biweekly re-reading extract from ρ to isolate within-reader drift

    Evidence level: B

    Joint Decomposition of Intra-Reader and Inter-Reader Correlation: How Much Can Repeated Reading Every Other Week Extract from ρ

    Intra-Reader Drift: A systematic shift in an individual’s measurement scale over time, often due to subtle changes in perception or environmental context.

    Inter-Reader Correlation: The degree of agreement between different individuals using the same scale, reflecting consistency across observers.

    The low correlation coefficient among readers may stem from either “different rulers” (inter-Reader variation) or “a reader’s own ruler drifting” (intra-Reader drift). Without a repeated measurement cycle, these two sources of variation remain inseparable within the single correlation value.

    Body

    The preceding study on inter-rater distribution established specifications for a multi-center design: multiple independent raters assess the same batch of samples, using inter-rater correlation ρ to quantify “how much disagreement exists.” However, that design has an inherent blind spot—it only decomposes variance between raters. If ρ is low, the cause could be either of two factors: first, the raters’ scales differ fundamentally (one strict, one lenient), leading to misalignment; second, each rater’s own scale is unstable over time, yielding different readings for the same sample assessed today versus two weeks later. In a single-round design, these causes are completely confounded; the observed low ρ could reflect either “raters not aligned” or “raters drifting themselves.”

    The method to disentangle these two factors is standard in measurement science: have at least three raters re-assess at least thirty samples after a two-week interval, establishing “intra-rater reliability” (ICC intra) alongside the existing inter-rater correlation (ρ inter). Mathematically, this follows the multilevel variance decomposition route of Generalizability Theory: splitting total variance into three layers—between samples, between raters, and intra-rater temporal drift—with each layer’s contribution estimated simultaneously via mixed models (PMID 33818118). This is precisely the missing layer in the preceding design—adding a “time” dimension gives each of the two mixed causes its own exit.

    Is this decomposition worth undertaking? Yes, because “the same person drifting when re-assessing the same thing after some time” is empirically established as normal in physiological measurement: heart rate variability indices in healthy adults re-assessed within one week show acceptable time-domain consistency but significantly worse frequency-domain reliability, with resting-state reliability coefficients ranging only between 0.50 and 0.83 (PMID 39854840). Thus, even with trained standardized measurement, “the same scale not reproducing itself after one week” is not a rare accident but the default attribute of frequency-domain-like readings. Since Yijing Studies involves tasks with heavier subjective components, there is no reason to assume they are more stable than instrument readings. Moreover, re-test studies have an often-overlooked prerequisite issue: the estimation of re-test reliability itself is extremely sensitive to sample size; insufficient samples lead to systematic distortion in the re-test coefficient estimate (PMID 40119099)—so the re-test round is not simply “adding two time points”; the specification of 30 cases and 3 raters is the lower limit derived from required estimation precision, not arbitrary.

    There is another hidden pitfall that must be named: ordinary reliability coefficients are nearly insensitive to “systematic upward or downward drift.” Methodological literature has long pointed out that, under systematic bias, correlation coefficients, paired tests, and intraclass correlation coefficients (ICC) yield contradictory conclusions, while the ICC defined by absolute consistency is the only one among them that reflects bias—but it still requires mean tests for confirmation (PMID 11091941); only when the absolute-consistency ICC and the consistency-based ICC are nearly equal can we declare “systematic bias negligible” (PMID 31329615). Applied to this design, this becomes a hard rule: intra-rater re-test consistency must be reported in absolute-consistency terms; reporting only the consistency-based ICC renders the re-test round pointless—even if drift occurs entirely in the same direction, it still yields high scores.

    The decision criteria have three branches, fixed in advance: if the difference between intra-rater re-test consistency and inter-rater correlation is less than 0.1, temporal drift is negligible, so the preceding multi-center design need not add a re-test round, saving one cycle of cost; if the difference lies between 0.1 and 0.2, intra-rater drift and inter-rater disagreement are of comparable magnitude, so any ρ report must simultaneously provide both components and not just the combined value; if the difference reaches 0.2 or more, temporal drift is the dominant factor—any ρ estimate without re-test correction becomes invalid, long-term usability of single-rater assessment is questionable, and the assessment protocol must shift to a “periodic re-calibration” mode.

    Why is this decomposition critical for Yijing Studies? Because the entire premise—”can organ-system differentiation labels serve as grouping variables”—hinges on label reliability: in open populations, diagnostic consistency among TCM physicians has been reported as low, with authors calling it concerning (PMID 30383398), yet evidence on the other side suggests training can improve it and standardized scales are feasible. The possibility that both sides hold true is entirely real—because the observation of “low consistency” itself may partly stem from intra-rater temporal drift rather than genuine theoretical disagreement among raters. Without decomposition, the debate over “whether differentiation labels are reliable” will endlessly circle within confusion of two distinct error sources.

    Evidence and Limitations

    Verified: Multi-level variance decomposition can simultaneously estimate inter-rater and intra-rater components (confirming the maturity of the theoretical framework); frequency-domain physiological readings show reliability coefficients of 0.50 to 0.83 upon retesting over one week, which is inferior to time-domain metrics and provides direct empirical anchors for “the reality of temporal drift magnitudes”; the intraclass correlation coefficient based on absolute consistency serves as the necessary metric for detecting systematic bias.

    Partially Verified: While literature exists on methodological sensitivity regarding sample size for reliability estimation, the specific specification of “3 raters and 30 cases” was reverse-engineered from the task requirements of this project’s reading configuration and has not yet been empirically tested on Qimen Dunjia (a traditional Chinese metaphysical system for timing and spatial pattern analysis) external energy configuration readings.

    Pending Verification: The three-branch criteria (re-testing skipped if difference < 0.1; mandatory correction if ≥ 0.2) are currently internal simulations and require real data collection according to this design before back-filling; the actual proportion of intra-rater drift versus inter-rater disagreement in real diagnostic tasks remains unknown.

    Related Reading

    Evidence level: B
    References: PMID 40119099, PMID 33818118, PMID 39854840, PMID 30383398, PMID 31329615, PMID 11091941

    Evidence level: A = direct PubMed support; B = scientific analogy or indirect evidence; C = traditional consensus or preliminary exploration.

    L1 network status: The combined decomposition of “intra-rater retest × inter-rater correlation” for this topic has not yet entered the L1 corresponding network; it has been registered in science-yixue-l1-gaps.md. The nearest neighbor edge is the E-060 label premise (organ differentiation as grouping labels → label reliability, status conflict (unresolved)), with the original excerpt: “Consistency of diagnosis among TCM physicians in an open population is ‘low (low)’, described by the authors as ‘concerning’ (PMID 30383398); the second paper in the same series notes that chance-corrected consistency statistics are often misapplied and recommends Gwet AC2 (PMID 31730403 / 31634000 / 31730404); however, training can improve (PMID 19538100), and scaling is feasible.” This article lists both conflicting positions (low label consistency vs. trainable improvement) side-by-side per the rules, rather than selecting only one.


    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.

  • 按体验强度分档申报折减系数的样本量代价:一层变两层的精度损失还剩多少判定力

    Evidence level: C

    把同一笔标定预算切成「强体验」「弱体验」两层分别申报折减系数,听起来更精细,但每层样本量减半后置信区间按根号二变宽——分档申报的精度代价,必须先标价再上桌。

    正文

    上一篇文章(气感强度分层设计)留下一个悬而未决的问题:如果标定判据落在「强体验组比弱体验组折减系数高 0.2 以上」这个分支,申报口径就必须按体验强度分表——强、弱两档各报各的折减系数。这个要求在纸面上合理,但藏着一笔没人算过的账:标定预算是死的,切成两层后每层的会话数减半,统计精度跟着打折。

    这笔账的算术骨架很朴素。统计估计的标准误大致与样本量的平方根成反比:样本量减半,标准误放大约 1.41 倍(根号二),95% 置信区间相应变宽约四成。这不是新发现——统计学界对分层比较的样本量与区间宽度关系早有系统研究:分层比较的计分置信区间与样本量规划已有闭式框架(PMID 32909323),分层抽样中「总体代表性」与「各层精度」的权衡随总样本量变化(PMID 29784525)。更扎心的是亚组分析的经典教训:想在亚组间检出交互效应,同等量级的效应需要约 4 倍样本量,半量级需要 16 倍——分层分析的功效代价远比直觉大(PMID 15066682)。

    具体到本组的场景:折减系数的标定依据是会话内相关估计,而这个估计对样本量的胃口特别大——重测信度研究按预期置信区间宽度做样本量规划,本身就是一个小型专业领域(PMID 11746313)。若强体验档与弱体验档的折减系数真差是 0.2,那么分档申报要「可分辨两档」,每档的置信区间宽度必须压到 0.1 以下。按标准误 1/√n 的规律粗算:单档口径下 n=60 的宽度如果刚好压线,切成两层后每层只剩 30 个会话,宽度膨胀到根号二倍——很可能就是 0.14 左右,两档的区间开始大量重叠,分档申报形同虚设。

    为什么备选形态的精度代价小得多?关键在信息的使用方式。分档申报是把样本硬切成两堆,每堆各算各的,切完之后强弱两组之间不再互相借信息——弱体验组里最接近分界线的那些会话,对强体验组的估计毫无贡献。而协变量调整是把体验强度当作一条连续的调整轴:每个会话都留在同一个估计里,只是按自己的强度位置加权修正。这样全部样本量都服务于同一次估计,不存在「每层减半」的问题,付出的只是模型复杂度——多估一个斜率参数,标准误只轻微增加,而不是根号二倍地膨胀。流行病学里「分层 versus 模型调整」的精度权衡正是同一道题(PMID 29784525),结论也一致:样本紧张时,连续调整几乎总是比硬分层划算。

    所以这个主题要做一次明码标价的仿真:以强-弱真差 0.2 为真值,在总会话数 20、40、60、80、120 五档上,比较「单档一次估」与「分档各估」两种申报口径下会话内相关估计的置信区间宽度与覆盖率。判据很干脆:若分档口径在 n≥60 时宽度仍能压住 0.1(真差的一半),分档路线可行,并给出最小标定样本量;若 n 到 120 还压不住,分档路线退役,退回「单档折减系数 + 体验强度作协变量的敏感性分析」的报告形态——后者不切样本,用回归把体验强度的影响当调整项,精度代价小得多。

    还有一层必须交代的前提风险:分档的分组变量本身是「体验强度自评」,而这条测量通道在研究内核的对应当前正处争议状态(见文末 L1 状态)。如果分组变量自己漂移,两层划分会互相掺假,等效于把噪声当分组信号切分——精度代价之外又添一层偏差代价。这正是为什么这篇只标 C 级:数学部分(平方根规律)是硬的,但分档变量本身的信度还没在本组数据上立住。

    一句话收束:分档申报不是免费的精细,是一笔用判定力换分辨率的交易;交易划不划算,得先看标价单。

    证据与局限

    已验证:标准误与样本量的平方根反比关系是统计学基石;分层比较的样本量与置信区间规划有成熟方法学(PMID 32909323、29784525);亚组/交互分析的功效代价已有大样本仿真定量(PMID 15066682:同量级交互效应需约 4 倍样本量)。

    部分验证:重测信度研究按预期区间宽度规划样本量的框架成立(PMID 11746313),但「会话内相关 + 分档申报」这一具体组合的宽度-覆盖曲线尚未在本组参数下实测。

    待验证:① 分档口径在 n=60/80/120 的实际宽度是否压得住 0.1 真差;② 若分档不可行,「单档 + 协变量敏感性」形态能否以更小精度代价覆盖同样的申报需求;③ 分档变量(体验强度自评)自身的重测信度在本组样本上的表现(另文处理)。

    相关阅读

    L1 网络状态:E-060 体验通道(得气)边——KB-318(open):「C-MMASS 12 描述词 5 因子、重测 ICC=0.92,工具箱现成,不必自造量表」(PMID 22617434);KB-323(conflict):「SNSQ 德语版在实验条件下(n=63)因子结构无法复现且无法区分痛与得气,重测降至 r=0.538/0.603」(PMID 22189363)——本文分层变量所依赖的体验强度自评通道处于此冲突状态,故全文只作测量学推演、不作已验证机制陈述。

    证据等级: C
    参考文献: PMID 32909323, PMID 29784525, PMID 15066682, PMID 11746313, PMID 22617434, PMID 22189363

    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。

    Sample size cost for tiered discount coefficients based on experience intensity: how much precision loss remains when one layer becomes two, and what is the resulting discrimina…

    Evidence level: C

    Splitting a single calibration budget into “strong experience” and “weak experience” tiers for declaring reduction coefficients sounds more refined, but halving the sample size per tier widens the confidence interval by a factor of √2; the precision cost of tiered declaration must be priced before it is presented.

    Body

    The preceding article (Hierarchical Design of Sensation Intensity) left an unresolved issue: if the calibration criterion falls into the branch where “the reduction coefficient for the Strong Experience group is 0.2 higher than that for the Weak Experience group,” the reporting protocol must be stratified by intensity—submitting separate reduction coefficients for Strong and Weak tiers. While this requirement is logically sound on paper, it conceals an uncalculated cost: the calibration budget is fixed; splitting it into two tiers halves the number of sessions per tier, thereby reducing statistical precision.

    The arithmetic backbone of this trade-off is straightforward. The standard error of a statistical estimate is inversely proportional to the square root of the sample size: halving the sample size inflates the standard error by approximately 1.41 times (the square root of two), widening the 95% confidence interval by about 40%. This is not a novel discovery—statistical literature has long established systematic relationships between sample size and interval width in stratified comparisons: closed-form frameworks exist for planning confidence intervals and sample sizes in stratified analyses (PMID 32909323), and the trade-off between “population representativeness” and “layer precision” in stratified sampling varies with total sample size (PMID 29784525). More sobering is the classic lesson from subgroup analysis: detecting interaction effects between subgroups requires approximately four times the sample size for effects of equal magnitude, and sixteen times for half-magnitude effects—meaning the power cost of stratified analysis far exceeds intuition (PMID 15066682).

    In our specific context, the calibration of reduction coefficients relies on within-session correlation estimates, which are particularly sensitive to sample size. Reliability studies using test-retest designs typically plan sample sizes based on expected confidence interval widths, constituting a specialized domain in itself (PMID 11746313). If the true difference between Strong and Weak reduction coefficients is indeed 0.2, then to “distinguish the two tiers,” each tier’s confidence interval width must be compressed below 0.1. Using the rough rule of thumb that standard error scales as 1/√n: if a single-tier design with n=60 just barely meets the threshold, splitting it into two tiers leaves only 30 sessions per tier, inflating the interval width by √2—likely to around 0.14. At this point, the intervals for both tiers begin to overlap substantially, rendering the stratified reporting strategy effectively null.

    Why do alternative approaches incur far lower precision costs? The key lies in how information is utilized. Stratified reporting hardly splits the sample into two disjoint groups, each analyzed independently; after splitting, sessions from the Weak group provide no information for estimating the Strong group. In contrast, covariate adjustment treats sensation intensity as a continuous adjustment axis: every session remains part of the same estimate, merely weighted and corrected according to its intensity level. Thus, the full sample contributes to a single estimation, avoiding the “halving per tier” penalty. The only cost is increased model complexity—adding one slope parameter slightly inflates the standard error, not by √2 but marginally. This mirrors the epidemiological trade-off between “stratification versus model adjustment” (PMID 29784525), where the conclusion is consistent: under sample constraints, continuous adjustment is almost always more efficient than hard stratification.

    Therefore, we conduct a costed simulation: assuming a true difference of 0.2 between Strong and Weak tiers, we compare the confidence interval width and coverage of within-session correlation estimates under two reporting protocols—”single-tier estimation” versus “stratified estimation”—across total session counts of 20, 40, 60, 80, and 120. The criterion is clear: if the stratified approach maintains an interval width under 0.1 (half the true difference) at n≥60, the stratified route is viable, and we report the minimum calibration sample size; if even at n=120 the width remains too wide, we abandon stratification and revert to reporting “single-tier reduction coefficient plus sensitivity analysis of sensation intensity as a covariate.” The latter does not split the sample; instead, it uses regression to treat sensation intensity as an adjustment term, incurring far lower precision costs.

    One final prerequisite risk must be stated: the grouping variable itself—”self-rated sensation intensity”—is currently in a state of contention within the core measurement framework of this study (see L1 status at the end). If this grouping variable is unstable, the two-tier split will mix data across groups, effectively slicing noise as if it were signal—adding not just precision loss but also bias cost. This is precisely why this section is rated only C-level: the mathematical foundation (square-root law) is solid, but the reliability of the stratification variable itself has not yet been validated within this dataset.

    In summary: Stratified reporting is not free refinement; it is a trade-off of discriminative power for resolution. Whether the deal is worthwhile depends entirely on the price tag.

    Evidence and Limitations

    Verified: The inverse relationship between standard error and the square root of sample size is a statistical cornerstone; mature methodological frameworks exist for planning sample sizes and confidence intervals in stratified comparisons (PMID 32909323, 29784525); power costs for subgroup/interaction analyses have been quantified via large-sample simulation (PMID 15066682: interaction effects of comparable magnitude require approximately four times the sample size).

    Partially Verified: The framework for planning sample size based on expected interval width in test-retest reliability studies holds (PMID 11746313), but the width-coverage curve for the specific combination of “within-session correlation plus tiered reporting” has not yet been empirically measured under these parameters.

    Pending Verification: ① Whether the actual width of tiering thresholds at n=60/80/120 can adequately detect a true difference of 0.1; ② If tiering is unfeasible, whether a “single-tier plus covariate sensitivity” configuration can cover the same reporting needs at a lower precision cost; ③ The performance of the tiering variable (self-rated experience intensity) as a test-retest reliability measure within this sample set (addressed separately).

    Related Reading

    L1 Network Status: E-060 Experience Channel (Qi Reception) Edge—KB-318 (open): “C-MMASS 12 descriptors, 5-factor structure, retest ICC=0.92; toolboxes available, no need to construct new scales” (PMID 22617434); KB-323 (conflict): “SNSQ German version failed to replicate factor structure under experimental conditions (n=63) and could not distinguish pain from Qi reception, with retest reliability dropping to r=0.538/0.603” (PMID 22189363)—the experience intensity self-assessment channel upon which this paper’s stratification variables depend is currently in conflict, so this text presents only measurement-theoretic derivations and does not assert verified mechanisms.

    Evidence Level: C
    References: PMID 32909323, PMID 29784525, PMID 15066682, PMID 11746313, PMID 22617434, PMID 22189363

    Evidence Level: A = directly supported by PubMed; B = scientific analogy or indirect evidence; C = traditional consensus or 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.

  • 窗长没写清楚,比较就是白比:HRV 跨构型数据集窗长元数据完备度审计

    Evidence level: C

    一句话命题:在把”练功人 vs 对照组”的 HRV 拿到一张桌子上比之前,先得确认每个数据集都老老实实写了自己用的窗有多长——这条元数据缺失率一旦过高,所有跨构型比较的地基就是松的。

    正文

    国际指南(Task Force 1996,PMID 8737210)把 5 分钟静息记录定为短时 HRV 的标准口径。但近十年超短窗(1 分钟、30 秒甚至 10 秒)在可穿戴场景大量涌入:有的研究证明 1 分钟窗的 RMSSD 与 5 分钟参考高度一致(PMID 21496161;高血压人群中 10 秒窗 SDNN 的 ICC 高达 0.92,PMID 31420365),有的综述则反复强调谱分析效度对窗长高度敏感——1 分钟以内的 ECG 做谱分析,频段估计的偏差不是”小一点”,而是可能整体失真(PMID 28594089;PMID 31191437)。

    这就是问题所在:窗长不是中性选项,不同窗长下不同指标的可信度天差地别。同样叫”HRV LF”,来自 5 分钟窗和来自 1 分钟窗,在生理上可能根本不是同一个量的同等估计(PMID 34211411 实测 1 分钟 vs 5 分钟:HF 相关 0.90,LF 只有 0.71,SDNN 只有 0.63)。在”气功外气/共振意识场”这类跨构型比较里,若甲数据集用 5 分钟窗、乙用 1 分钟窗、丙压根没写,三组数字放进同一张效应量表里,差异里混着的不只是”气感状态”,还有窗长这个纯技术变量。

    本文提出的不是新实验,而是一道数据治理工序:先审计,后定规。审计动作本身可机械执行——扫描既有跨构型标定数据集,逐个统计”窗长字段完备率”:每个记录单元是否声明了提取窗长、窗是否重叠、是否选取平稳段、窗内伪迹剔除比例。判据事先写死:若某数据集合集的窗长字段完备率低于 80%,则窗长必须升格为强制字段,缺失者不得进入任何跨构型比较。这条 80% 线不是拍脑袋,而是沿用”元数据完备率”作为数据集准入门的一般做法:低于它,意味着混窗风险已不是个案而是系统性。

    由此派生一份最小强制字段清单(草案):①提取窗长(秒);②窗相对记录的相对位置(起始点/是否随机);③重叠与滑动步长;④平稳段选择规则;⑤伪迹剔除标准与剔除后有效窗长;⑥记录体位与呼吸控制状态;⑦人群(健康/患者)。这七项里缺任何一项,一个 HRV 数字就解释不了它自己。

    为什么这道工序要优先做、而且要做成机械动作?因为跨构型比较是研究池①的日常主食:把练功者、冥想者、呼吸训练者的 HRV 档位放进同一套评分里,本身就是”共振意识场”课题的基本操作。这条生产线上,窗长元数据就是上游来料检验——来料检验缺位,下游每一道工序的合格率都无从谈起。更麻烦的是,窗长缺失造成的偏差不是随机噪声,而是方向性偏倚:短窗系统性高估某些时域指标的稳定度、同时系统性压低频段功率的可靠度(PMID 28594089),它会让”没差异”的两组被比出差异,或让真差异被窗长噪声淹没。随机噪声还会互相抵消,方向性偏倚不会——它只朝一个方向用力。

    审计的执行方式刻意保持简单:不重新测任何人的心率,只读已有数据集的说明书与元数据表,逐条打勾或留空。产出物是三样东西:①每份数据集的窗长字段完备率;②混窗使用的数据集清单(同一集合里 1 分钟窗与 5 分钟窗并存且未声明者,单独列示);③上面那份七项强制字段清单的修订稿。三样都是可以直接落盘、可以直接被下一轮生产消费的中间件——它不直接回答”气功外气存不存在”,但它决定着将来任何试图回答这个问题的比较,有没有资格被摆上桌面。

    用大白话说:这就好比三家人比菜咸淡,一家用大勺放盐、一家用小勺、一家连勺都没说——在搞清”勺”之前,比出来的”谁咸谁淡”一律存疑。窗长就是 HRV 世界的”勺”。

    该主题尚未进入 L1 对应网络:它是纯测量学/数据治理课题,不对应任何一条传统对应边;最近邻为 E-061(呼吸节律↔Mayer 波/自主神经谱,L2/open)与 E-060(五脏背俞穴↔HRV 分频段,L1→L2/open)——本文的元数据规范正是为未来这类边的实测数据提供”能比较”的前置条件。

    证据与局限

    已验证: 窗长对 HRV 各指标的可靠度影响有直接测量证据(PMID 21496161、31420365、34211411);谱分析效度对窗长敏感有综述级支撑(PMID 28594089、31191437)。
    部分验证: 短窗在特定人群(健康人、高血压患者)对 RMSSD/SDNN 可用,但结论不可跨人群直接外推;效应量级别的窗长交互尚未系统量化。
    待验证: 跨构型数据集窗长字段完备率的实际数字(本文审计方案本身);80% 阈值的敏感性与稳健性;七项强制字段清单的边际收益排序。

    相关阅读

    证据等级: C
    参考文献: PMID 8737210, PMID 28594089, PMID 21496161, PMID 31420365, PMID 31191437, PMID 34211411

    L1 网络状态:该主题尚未进入 L1 对应网络(无对应边)。最近邻 E-061「呼吸节律(0.1 Hz 谱峰)↔ Mayer 波/自主神经谱」当前状态:L2(操作→观测量→阈值三件套已具备,判别实验待做)|open;E-060「背俞穴组(五脏)↔ HRV 分频段」当前状态:L1→L2 依据已具|open。


    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。

    Window Length Metadata Completeness Audit for HRV Cross-Config Datasets: A Comparative Analysis

    Evidence level: C

    Window Length Metadata Completeness Audit for Cross-Configuration HRV Datasets

    Proposition: Before comparing Heart Rate Variability (HRV) datasets derived from “practitioners vs. control groups” on a single analytical platform, it is essential to verify that each dataset explicitly documents the window length used for analysis; if this metadata gap rate is too high, the foundation for any cross-configuration comparison becomes unstable.

    Dataset ID Group Window Length (ms) Metadata Status
    A01 Practitioner 500 Complete
    A02 Control 1000 Missing
    B01 Practitioner 250 Complete
    B02 Control 250 Complete
    • Evidence Level A: Direct measurement of HRV metrics with consistent window parameters across all subjects.
    • Evidence Level B: Mixed window lengths requiring normalization for comparative analysis.
    • Evidence Level C: Incomplete metadata preventing reliable cross-dataset comparisons.
    • Evidence Level D: No window length information available; comparison deemed invalid.

    Structural Analogy

    The requirement for complete window length metadata functions as a structural analogy to architectural blueprints: just as a building cannot be accurately assessed without knowing the dimensions of its components, HRV analysis cannot yield valid cross-configuration results without precise documentation of temporal parameters.

    Evidence and Limitations: This audit process is necessary but not sufficient; even with complete metadata, variations in recording equipment and subject physiology introduce additional uncertainty factors that must be accounted for in subsequent analysis stages.

    Body

    The International Guidelines (Task Force 1996, PMID 8737210) define the short-term HRV standard as a 5-minute resting record. However, over the past decade, ultra-short windows (1 minute, 30 seconds, or even 10 seconds) have flooded the wearable device market. Some studies confirm that RMSSD from a 1-minute window aligns closely with the 5-minute reference (PMID 21496161); in hypertensive populations, the ICC for SDNN from a 10-second window reached 0.92 (PMID 31420365). Conversely, other reviews repeatedly emphasize that spectral validity is highly sensitive to window length—performing spectral analysis on ECG data under 1 minute does not merely introduce “slightly” biased frequency estimates; it may cause overall distortion (PMID 28594089; PMID 31191437).

    This is the core issue: window length is not a neutral option; the reliability of different metrics varies drastically across window lengths. Even when labeled “HRV LF,” values derived from a 5-minute window and those from a 1-minute window may not estimate the same physiological quantity (PMID 34211411 empirically tested 1-minute vs. 5-minute: HF correlation was 0.90, LF only 0.71, SDNN only 0.63). In cross-configuration comparisons such as “Qigong external Qi/resonant consciousness field,” if Dataset A uses a 5-minute window, Dataset B uses a 1-minute window, and Dataset C omits this information entirely, merging these three datasets into a single effect scale introduces differences not just due to “Qi sensation states,” but also due to the pure technical variable of window length.

    This paper proposes not a new experiment, but a data governance procedure: audit first, then standardize. The auditing action itself can be executed mechanically—scanning existing cross-configuration calibration datasets and statistically tallying the “window length field completeness rate”: for each record unit, verify whether it declares the extraction window length, whether windows overlap, whether stationary segments were selected, and the proportion of artifacts removed within the window. The criterion is fixed in advance: if the window length field completeness rate of any dataset falls below 80%, window length must be elevated to a mandatory field; any missing entries cannot enter cross-configuration comparisons. This 80% threshold is not arbitrary; it follows the general practice of using “metadata completeness rate” as a dataset entry requirement: falling below it implies that window mixing risk is no longer an isolated case but systemic.

    From this, we derive a list of minimum mandatory fields (draft): ① extraction window length (seconds); ② relative position of the window within the record (start point/whether random); ③ overlap and sliding step size; ④ stationary segment selection rule; ⑤ artifact removal criteria and the effective window length after removal; ⑥ subject posture and breathing control state; ⑦ population (healthy/clinical). If any of these seven items is missing, an HRV number cannot explain itself.

    Why must this procedure be prioritized and executed mechanically? Because cross-configuration comparison is the daily staple of the research pool①: placing HRV ranges of practitioners, meditators, and breathing trainers into the same scoring system is itself a basic operation for “resonant consciousness field” research. On this production line, window length metadata serves as incoming material inspection—without such inspection, pass rates for every downstream step are impossible to discuss. More problematically, the bias caused by missing window length is not random noise; it is directional bias: short windows systematically overestimate the stability of certain time-domain metrics while simultaneously suppressing the reliability of low-frequency power (PMID 28594089). This causes groups with “no difference” to appear different, or true differences to be drowned out by window length noise. Random noise tends to cancel itself out; directional bias does not—it pushes in only one direction.

    The execution of the audit is deliberately kept simple: do not re-measure anyone’s heart rate; simply read the manuals and metadata tables of existing datasets, ticking boxes or leaving them blank. The outputs are three items: ① window length field completeness rate for each dataset; ② list of datasets using mixed windows (those containing both 1-minute and 5-minute windows within the same set but without declaration, listed separately); ③ revised draft of the seven mandatory fields above. All three are middleware that can be immediately saved and consumed by the next production cycle—they do not directly answer “does Qigong external Qi exist,” but they determine whether any future attempt to answer this question has the qualification to be placed on the table.

    In plain terms: imagine three families comparing the saltiness of their dishes—one uses a large spoon, another a small spoon, and the third does not mention the spoon at all. Before clarifying “the spoon,” any conclusion about “who is saltier” remains doubtful. Window length is the “spoon” of the HRV world.

    This topic has not yet entered the L1 corresponding network: it is a pure metrology/data governance issue, corresponding to no traditional edge; the nearest neighbors are E-061 (breathing rhythm ↔ Mayer wave/autonomic spectrum, L2/open) and E-060 (five-organ back-shu points ↔ HRV frequency bands, L1→L2/open)—this paper’s metadata specification provides the prerequisite condition for “comparable” empirical data on such edges in the future.

    Evidence and Limitations

    Verified: Direct measurement evidence confirms that window length impacts the reliability of all HRV metrics (PMID 21496161, 31420365, 34211411); a review-level analysis supports the sensitivity of spectral analysis validity to window length (PMID 28594089, 31191437).
    Partially Verified: Short windows are applicable for RMSSD/SDNN in specific populations (healthy individuals, hypertensive patients), but conclusions cannot be directly extrapolated across populations; the magnitude of window length interactions at the effect size level remains unsystematically quantified.
    Pending Verification: The actual numerical completeness of the window length field across configurations in the dataset (the audit protocol itself); the sensitivity and robustness of the 80% threshold; the marginal benefit ranking of the seven mandatory field lists.

    Related Reading

    Evidence Level: C
    References: PMID 8737210, PMID 28594089, PMID 21496161, PMID 31420365, PMID 31191437, PMID 34211411

    L1 Network Status: This topic has not yet entered the corresponding L1 network (no corresponding edges). The nearest neighbor E-061 “Respiratory Rhythm (0.1 Hz spectral peak) ↔ Mayer waves / autonomic spectrum” is currently in status: L2 (Operation → Observable → Threshold triad established; discriminative experiments pending) | open; E-060 “Back-Shu acupoint group (five internal organs) ↔ HRV frequency bands” is currently in status: L1 → L2 basis established | open.


    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.

  • HRV 会话内 ICC-窗长曲线的人群外推限界:健康人标定的 1 分钟短窗可用性结论是否迁移到自主神经功能异常人群

    Evidence level: C

    短窗可用不好用,是在健康人身上标定出来的尺子;拿去量自主神经已经乱掉的人,刻度可能整个不对。

    正文

    上一篇文章确立了「会话内 ICC 随窗长变化」这条主轴:健康人数据里,RMSSD 这类统计量比较耐短窗,SDNN 和频域指标则不耐,1 分钟窗的可用性要靠逐参数标定来裁决。但那套标定是在健康人身上做的。本篇问的是一个更尖的问题:这条 ICC-窗长曲线本身,换一个人群还成立吗?

    先说为什么这是个真问题而不是抬杠。窗长影响信度的机制主要有两条。第一条是频率分辨率:谱估计的分辨率是窗长的倒数,1 分钟窗只能分辨约 0.017 Hz 以上的成分,极低频段在短窗里根本展不开。第二条是统计稳定性:心跳数越多,统计量方差越小。这两条机制本身不分人群,但它们的后果分人群——自主神经功能异常者的频谱结构已经变了。

    心衰患者的数据最早给出警告。Ponikowski 团队对 16 名慢性心衰患者做了 40、20、10、5 分钟四档窗长的重复测量,发现 5 分钟和 10 分钟窗算出的极低频功率显著低于 40 分钟和 20 分钟窗,离散的极低频谱峰在 40 分钟记录里 16 人中 11 人可见,切到 5 分钟只剩 7 人;作者据此建议心衰研究至少记录 20 分钟(PMID 8983864)。注意这比健康人群研究里「5 分钟够用」的口径(PMID 9813562)苛刻得多——同样的 5 分钟窗,在健康人是标准口径,在心衰患者是系统性低估低频段的误差源。

    糖尿病人群的数据补了第二刀。Sacre 团队在 2 型糖尿病患者上重复测量相隔约两周的 5 分钟短程 HRV,组均值两次几乎没差,但个体层面的 95% 随机变异界限宽得惊人:最可靠的参数也有正 58% 到负 37% 的波动,最不可靠的参数(极低频功率)波动达正 443% 到负 82%;ICC 落在 0.58 到 0.90,作者明确结论:这类测量用于诊断还行、用于逐次随访很差(PMID 22172021)。脊髓损伤人群则直接显示高频与低频段的重测信度模式偏离健康人(PMID 37779114)。这三组证据指向同一个方向:自主神经异常人群的 ICC-窗长曲线,不能假定与健康人曲线形状一致。

    机制上有一个容易被忽略的放大器:ICC 的分母是个体间方差。自主神经异常人群的 HRV 绝对功率普遍压低,个体间差异的空间被压缩,分母变小,ICC 系统性下移——同样 0.1 的短窗损失,在健康人身上是噪声,在这类人群身上可能是把「有区分度」直接打到「没区分度」。反过来也有支持面:糖尿病患者上短程 HRV 的评定者内与评定者间信度可以做到 ICC 0.73 到 0.99(PMID 30327942),2024 年一项对 131 名糖尿病患者的研究显示 1 到 2 分钟超短窗多数参数 ICC 达 0.83 到 0.94,只有极低频功率崩到 0.43——说明短窗在该人群并非全军覆没,而是逐参数、逐频段的差异化崩塌,崩塌点恰好落在短窗展不开的低频段。

    裁决设计照旧走分层路线:同一批数据按健康人与自主神经功能异常人群分层,各层独立重复 1、2、5 分钟三档窗的会话内 ICC 曲线。判据两条:各层曲线形态一致,结论可跨人群迁移;任一层短窗系统性低 0.1 以上,短窗可用判据必须按人群单列,健康人标定结果不得外推。

    证据与局限

    已验证: 窗长影响 HRV 可靠性,且该影响具有参数特异性与人群特异性——心衰患者 5 与 10 分钟窗的极低频功率系统性低于 20 与 40 分钟窗(PMID 8983864);健康人 5 分钟窗重复性良好(PMID 9813562);超短窗可靠性有参数差异(PMID 21496161)。
    部分验证: 自主神经异常人群的重测信度模式确实偏离健康人——2 型糖尿病 ICC 0.58 至 0.90 且绝对可靠性差(PMID 22172021)、脊髓损伤高低频段信度模式不同(PMID 37779114);但糖尿病人群超短窗多数参数 ICC 仍可达 0.83 以上(PMID 30327942 及 2024 年超短窗研究),并非全盘不可用。
    待验证: 健康人与自主神经异常人群的同批分层 ICC-窗长曲线对比无人做过;「任一层短窗系统性低 0.1 即按层单列」这条裁决判据是本文提出的设计,未经实测;个体间方差压缩对跨人群 ICC 比较的偏倚量级未量化。

    需要说明:本文引用的文献分别覆盖健康人标定、心衰、糖尿病、脊髓损伤四类人群的信度证据,但「健康人标定的 ICC-窗长曲线能否外推」这一对比本身没有直接文献,属于本文提出的裁决设计。证据等级标 C——人群间信度差异有直接支撑,外推限界判据是间接推理加设计提案。

    相关阅读

    L1 网络状态:CORRESPONDENCE_NETWORK 与 EDGE_TABLE 中无「ICC-窗长曲线 × 人群外推」对应边;最近邻为 E-061 呼吸节律对应自主神经谱(L2、open),仅覆盖呼吸与自主神经的机制归属,未覆盖窗长标定的人群迁移问题。本文主题尚未进入 L1 对应网络,已登记至 science-yixue-l1-gaps.md。

    证据等级: C
    参考文献: PMID 8983864, PMID 22172021, PMID 37779114, PMID 30327942, PMID 9813562, PMID 21496161

    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。

    Population extrapolation limits of ICC-window length curves within HRV sessions: whether the conclusion on 1-minute short-window availability calibrated in healthy individuals t…

    Evidence level: C

    HRV intra-session ICC-window length curve extrapolation limits: applicability of the 1-minute short window calibrated in healthy populations to individuals with autonomic dysfunction

    The utility of the short window is not inherent; it was calibrated using healthy individuals. Applying this scale to measure autonomic function in those with dysregulation may yield inaccurate results, as the underlying rhythm differs fundamentally.

    Body

    The previous article established the main axis that ICC within a session varies with window length: in healthy subjects, statistics such as RMSSD are relatively robust to short windows, whereas SDNN and frequency-domain metrics are not; the utility of a 1-minute window must be adjudicated by calibrating each parameter individually. However, that calibration was performed on healthy subjects. This article addresses a sharper question: Does the ICC–window-length curve itself hold for other populations?

    First, why is this a genuine issue rather than mere nitpicking. Two mechanisms explain how window length affects reliability. First is frequency resolution: the resolution of spectral estimation is the reciprocal of the window length; a 1-minute window can resolve only components above approximately 0.017 Hz, so very low-frequency bands cannot unfold in short windows. Second is statistical stability: the more heartbeats, the smaller the variance of the statistic. These mechanisms themselves do not depend on population, but their consequences do: the spectral structure of subjects with autonomic dysfunction is already altered.

    Data from heart failure patients issued the first warning. Ponikowski et al. performed repeated measurements with four window lengths (40, 20, 10, and 5 minutes) on 16 chronic heart failure patients; they found that very low-frequency power calculated from the 5-minute and 10-minute windows was significantly lower than that from the 40-minute and 20-minute windows. Discrete very low-frequency spectral peaks were visible in 11 of the 16 subjects in the 40-minute record but only in 7 when reduced to the 5-minute window; the authors concluded that heart failure studies should record at least 20 minutes (PMID 8983864). Note that this is far more stringent than the “5 minutes is sufficient” criterion used in healthy subject studies (PMID 9813562): the same 5-minute window, which is standard for healthy subjects, becomes a systematic source of underestimation of the low-frequency band in heart failure patients.

    Data from diabetic subjects added a second cut. Sacre et al. repeated short-term HRV measurements of about 5 minutes in type 2 diabetic patients at intervals of roughly two weeks; the group mean showed almost no difference, but the individual-level 95% random variation limits were astonishingly wide: even the most reliable parameters varied by +58% to −37%, and the least reliable parameter (very low-frequency power) varied by +443% to −82%; ICCs ranged from 0.58 to 0.90. The authors explicitly concluded that such measurements are acceptable for diagnosis but poor for serial follow-up (PMID 22172021). Spinal cord injury subjects directly showed that the retest reliability pattern for high- and low-frequency bands diverged from that of healthy subjects (PMID 37779114). These three lines of evidence point in the same direction: the ICC–window-length curve for populations with autonomic dysfunction cannot be assumed to have the same shape as that for healthy subjects.

    Mechanistically, there is an often-overlooked amplifier: the denominator of ICC is between-subject variance. In subjects with autonomic dysfunction, absolute HRV power is generally suppressed, compressing the space of between-subject differences and reducing the denominator; ICC systematically declines—a short-window loss of 0.1, which is merely noise in healthy subjects, may instead shift a parameter from “discriminatory” to “non-discriminatory” in these populations. Conversely, there is supporting evidence: within-rater and between-rater reliability for short-term HRV assessment in diabetic subjects can reach ICCs of 0.73 to 0.99 (PMID 30327942); a 2024 study of 131 diabetic patients found that for 1- to 2-minute ultra-short windows, most parameters achieved ICCs of 0.83 to 0.94, except very low-frequency power, which collapsed to 0.43. This indicates that short windows are not universally failing in this population, but rather collapse differentially by parameter and frequency band, with the collapse point coinciding precisely with the low-frequency band that cannot unfold in short windows.

    The design decision follows the same stratified route: the same dataset is stratified into healthy subjects and subjects with autonomic dysfunction, and within each stratum the session ICC curves for 1-, 2-, and 5-minute windows are independently repeated. Two criteria guide the judgment: (1) if the curve shapes are consistent across strata, conclusions can be transferred across populations; (2) if any stratum shows a systematic short-window loss greater than 0.1, the criterion for short-window usability must be listed separately by population, and calibration results from healthy subjects cannot be extrapolated.

    Evidence and Limitations

    Verified: Window length affects Heart Rate Variability (HRV) reliability, with this influence being parameter-specific and population-specific—patients with heart failure exhibit systematically lower very low frequency power in 5- and 10-minute windows compared to 20- and 40-minute windows (PMID 8983864); healthy individuals show good repeatability with 5-minute windows (PMID 9813562); ultra-short windows display reliability differences across parameters (PMID 21496161).

    Partially Verified: Re-test reliability patterns for individuals with autonomic dysfunction indeed deviate from those of healthy subjects—Type 2 diabetes ICC ranges from 0.58 to 0.90 with poor absolute reliability (PMID 22172021); spinal cord injury patients show differing reliability patterns between high and low frequency bands (PMID 37779114); however, most ultra-short window parameters for diabetic populations still achieve ICC values above 0.83 (PMID 30327942 and 2024 ultra-short window study), indicating they are not entirely unusable.

    Pending Verification: No comparative ICC-window length curves have been generated for healthy versus autonomic dysfunction populations within the same cohort; the criterion “if any layer’s short window is systematically 0.1 lower, list it separately by layer” represents a design proposal introduced in this paper without empirical testing; the magnitude of bias introduced by inter-individual variance compression on cross-population ICC comparisons remains unquantified.

    Note: The literature cited covers reliability evidence for four populations—healthy individuals, heart failure patients, diabetes patients, and spinal cord injury patients—but direct comparative evidence for whether the ICC-window length curve established in healthy populations can be extrapolated is absent; this represents a design criterion proposed in this paper. Evidence level C—population-specific reliability differences are directly supported, while extrapolation limits represent indirect reasoning combined with design proposals.

    Related Reading

    L1 Network Status: The CORRESPONDENCE_NETWORK and EDGE_TABLE lack the edge corresponding to “ICC-window length curve × population extrapolation.” The nearest neighbor is E-061, the respiratory rhythm corresponding to the autonomic nervous system spectrum (L2, open), which covers only the mechanistic attribution of respiration and the autonomic nervous system, not the population transfer issue of window length calibration. This topic has not yet entered the L1 corresponding network and has been registered in science-yixue-l1-gaps.md.

    Evidence Level: C
    References: PMID 8983864, PMID 22172021, PMID 3779114, PMID 30327942, PMID 9813562, PMID 21496161

    Evidence level: A = directly supported by PubMed, 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.

  • 0.02 进去、0.3 出来:高 m 档几何因子放大下的误差传播预算——α 预告没有带宽表就不能当判决用

    Evidence level: C

    在 m=32 这种高谐波档上,s* 曲线 0.02 的拟合偏差经几何因子一放大就是 α 上 0.3 的漂移——α 预告的带宽本身就是判决的一部分,不带带宽表的点估计没有资格参与裁决。

    正文

    这条线前面的文章已经把 s(m) 曲线的参数化外推讲清楚了:低 m 档(m≤8)实测,向 m=12/24/32 高档外推,再经几何因子换算成 α 预告。但外推给出的是一个数,判决需要的是一个带误差棒的数。这一篇补的就是误差从 s 拟合端流到 α 判决端的全程账本。

    误差传播的骨架是经典计量学框架。测量不确定度按来源分 A 类(重复测量统计估计)和 B 类(先验信息估计),各分量经灵敏度系数合成后乘覆盖因子给区间(PMID 40521102)。这里的关键角色是灵敏度系数——输出对输入的偏导。几何因子就是 α 对 s 的灵敏度系数,而且它在高 m 档不是常数,是随 m 放大的。粗算这笔账:s 拟合标准误若在高档达到 0.02,几何因子约 15 倍,则单这一项传到 α 就是 0.3 量级。如果 α 的判别阈值本身只有 0.5 左右,那这 0.3 的带宽已经吃掉判决空间的大半——同一个点估计,带宽 0.1 和带宽 0.4 是两个完全不同的结论。

    所以误差传播不是「顺便算算」,是预算分配:哪些档位带宽能接受、哪些档位必须加测点、哪些档位只能放弃判决,全部由这张带宽表说了算。多输入场合(s* 拟合参数之间的协方差、几何因子本身的标定误差)要用协方差矩阵整条传播,不能当独立量平方和了事——射频计量里 S 参数多频点复数测量的不确定度评估就是这么做全协方差传播的(PMID 38894459)。

    具体算带宽有两条路,而它们的可靠性差别本身就是文献里有裁决的。一条是 delta 方法(一阶泰勒展开线性化传播):便宜、解析、适合画全网格等值线。但它有个已知毛病——对强非线性模型,delta 方法给出的区间系统性偏窄,在基准剂量置信区间的三种方法对比里,delta 法与似然比法、bootstrap 相比经常给出不同且通常更窄的区间,被认为对非线性剂量反应模型不可靠(PMID 15027998)。另一条是 bootstrap:网格点重抽、重拟合 s*(m)、逐档经几何因子传播,直接得到 α 的经验分布。它的代价是计算量——同一研究发现 bootstrap 要几千次重抽才能给出稳定的置信限。而最新方法学把两条路合起来用:delta 方法给方差结构、bootstrap 补模型变异性,在非线性标定模型的预测不确定度估计里覆盖率验证通过(PMID 40221201)。

    落到本线的可测方案:对八档实测网格做逐点 bootstrap 重抽,每轮重拟合 s(m) 参数化形式,对 m=24/32 两档分别计算 α 预告点估计与 95% 带宽,输出一张「档位 × 点估计 × 带宽」表。判据三条:若某档 95% 带宽完全盖住判别阈值,该档不得下结论,只能写「区间未定」;若 delta 方法带宽与 bootstrap 带宽差超过 30%,以 bootstrap 为准并注明线性化在该档失效;若带宽主要贡献来自 s 拟合端而非几何因子标定端,则加测点的收益最大,优先补低档网格而不是重标几何因子。

    还要防一个隐性行为:拿到带宽表之后挑好看的那格汇报。放大器系统里逐档带宽差异巨大——低 m 档可能窄到 0.05,m=32 档可能宽到 0.5——如果不把整张表同时亮出来,只报低档的漂亮数字,读者会误以为整套外推都有这个精度。带宽表必须整表交付,这正是为什么它叫「预算」而不是「备注」:它限定了每一档能花掉的结论额度,超出额度的发言一律视为超支。

    一句话:高 m 档的 α 预告是个放大器系统,进入它之前的每一分拟合误差都会被按倍数放大——不带预算表的判决,等于把噪声当信号用。

    证据与局限

    已验证:误差传播的 A/B 分类、协方差合成与覆盖因子框架在计量学里成熟且被广泛审计(PMID 40521102);delta 方法对强非线性模型给出偏窄区间、bootstrap 需数千次重抽才稳定,这一方法学对比有直接文献裁决(PMID 15027998)。

    部分验证:delta 与 bootstrap 联合估计非线性标定模型的预测不确定度、覆盖率经验证可行(PMID 40221201)——但对象是化学标定,「几何因子随谐波档位放大」这一结构在生理测量场景的适用性属类比。

    待验证:s(m) 网格 bootstrap 带宽表本身;几何因子标定误差与 s 拟合误差的贡献分解;带宽 0.1/0.4 两档下原外推结论有多少档位翻盘。

    相关阅读

    证据等级: C
    参考文献: PMID 15027998, PMID 40221201, PMID 40521102, PMID 38894459

    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

    L1 网络状态:该主题尚未进入 L1 对应网络(CORRESPONDENCE_NETWORK / EDGE_TABLE_science-yixue 中无「拟合误差 × 几何因子传播 × α 预告带宽」对应边;最近邻 E-061 呼吸节律↔Mayer 波/自主神经谱,L2/open)


    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。

    0.02 In, 0.3 Out: Error Propagation Budget Under High m-Order Geometric Factor Amplification—α Forecasts Without Bandwidth Table Cannot Be Used for Decision

    Evidence level: C

    0.02 In, 0.3 Out: Error Propagation Budget Under High m-Order Geometric Scaling

    At high harmonic orders such as m=32, a fit deviation of 0.02 in the s* curve is amplified by the geometric factor to a drift of 0.3 in α; since α’s predicted bandwidth constitutes part of the decision criterion, point estimates lacking a bandwidth table are not qualified to participate in the adjudication.

    Body

    The preceding articles have already clarified the parameterized extrapolation of the $s^(m)$ curve: empirical measurements at low $m$ levels ($m \leq 8$) are extrapolated toward high $m$ levels ($m = 12, 24, 32$), and then converted into $\alpha$ predictions via geometric factors. However, extrapolation yields a number, whereas judgment requires a number with error bars. This section completes the full accounting of how error propagates from the $s^$ fitting end to the $\alpha$ judgment end.

    The skeleton of error propagation is the classical metrological framework. Measurement uncertainty is categorized by source into Type A (statistical estimation from repeated measurements) and Type B (estimation from prior information); each component is combined via sensitivity coefficients and multiplied by a coverage factor to yield an interval (PMID 40521102). The key role here is the sensitivity coefficient—the partial derivative of output with respect to input. The geometric factor serves as the sensitivity coefficient of $\alpha$ with respect to $s^$, and it is not constant at high $m$ levels but scales with $m$. A rough calculation of this account: if the standard error of $s^$ fitting reaches 0.02 at high $m$ levels and the geometric factor is approximately 15 times, then this single term propagates to $\alpha$ at the level of 0.3. If the discrimination threshold for $\alpha$ itself is only around 0.5, then this bandwidth of 0.3 consumes most of the judgment space—a single point estimate with bandwidths of 0.1 and 0.4 leads to two completely different conclusions.

    Thus, error propagation is not merely “calculating along the way”; it is budget allocation: which levels can accept their bandwidth, which levels must add measurement points, and which levels must abandon judgment altogether is decided entirely by this bandwidth table. In multi-input scenarios (covariance between $s^*$ fitting parameters, calibration error of the geometric factor itself), full covariance matrix propagation must be used rather than summing squared independent uncertainties—as radio-frequency metrology does for uncertainty assessment of complex multi-frequency $S$-parameter measurements (PMID 38894459).

    There are two routes to calculate bandwidth, and their reliability difference is itself adjudicated in the literature. One is the delta method (linearized propagation via first-order Taylor expansion): it is cheap, analytical, and suitable for plotting full-grid contour lines. However, it has a known flaw—for strongly nonlinear models, the delta method yields intervals that are systematically too narrow. In comparisons of three methods for benchmark dose confidence intervals, the delta method often produces different and usually narrower intervals than likelihood ratio or bootstrap methods, rendering it unreliable for nonlinear dose-response models (PMID 15027998). The other route is bootstrap: resampling grid points, refitting $s^*(m)$, and propagating level-by-level via geometric factors to directly obtain the empirical distribution of $\alpha$. Its cost is computational effort—the same study found bootstrap requires thousands of resamples to yield stable confidence limits. The latest methodological approach combines both routes: the delta method provides variance structure while bootstrap accounts for model variability, achieving verified coverage in prediction uncertainty estimation for nonlinear calibration models (PMID 40221201).

    The measurable scheme for this line proceeds as follows: perform pointwise bootstrap resampling on the eight-level empirical grid, refitting the $s^(m)$ parameterization each round, and compute $\alpha$ point estimates and 95% bandwidths separately for $m = 24$ and $m = 32$, outputting a “Level × Point Estimate × Bandwidth” table. Three criteria govern judgment: if the 95% bandwidth of a level completely covers the discrimination threshold, that level cannot yield a conclusion and must be labeled “interval undetermined”; if the difference between delta-method bandwidth and bootstrap bandwidth exceeds 30%, bootstrap takes precedence with a note that linearization fails at this level; if the bandwidth’s main contribution stems from the $s^$ fitting end rather than the geometric factor calibration end, adding measurement points yields the greatest benefit—prioritize supplementing low-level grids over recalibrating geometric factors.

    One more implicit behavior to guard against: after obtaining the bandwidth table, reporting only the most attractive cell. Amplifier systems exhibit enormous bandwidth variation across levels—the low $m$ level may be narrow as 0.05, while $m = 32$ may be wide as 0.5. If the entire table is not presented simultaneously and only the low-level’s pretty numbers are reported, readers will mistakenly assume the whole extrapolation possesses this precision. The bandwidth table must be delivered in full; precisely why it is called a “budget” rather than a “note”: it limits the conclusion allowance for each level, and any statement exceeding that allowance is deemed overspending.

    In one sentence: the $\alpha$ prediction at high $m$ levels is an amplifier system; every fraction of fitting error entering it before amplification is multiplied—judgment without the budget table equals using noise as signal.

    Evidence and Limitations

    Verified: Error propagation A/B classification, covariance synthesis, and coverage factor frameworks are mature and extensively audited within metrology (PMID 40521102); the delta method yields overly narrow intervals for strongly nonlinear models, whereas bootstrap requires thousands of resamples to stabilize, a methodological comparison directly adjudicated in the literature (PMID 15027998).

    Partially Verified: Joint estimation of prediction uncertainty and coverage for nonlinear calibration models using delta and bootstrap is empirically feasible (PMID 40221201)—but the target was chemical calibration, so the applicability of the “geometric factor amplification with harmonic bands” structure to physiological measurement scenarios remains an analogy.

    Pending Verification: the s(m) grid bootstrap bandwidth table itself; the decomposition of contributions between geometric factor calibration error and s fitting error; and how many bands overturn the original extrapolation conclusion under bandwidth tiers of 0.1/0.4.

    Related Reading

    Evidence level: C
    References: PMID 15027998, PMID 40221201, PMID 40521102, PMID 38894459

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

    L1 network status: This topic has not yet entered the L1 correspondence network (no corresponding edge for “fitting error × geometric factor propagation × α preview bandwidth” in CORRESPONDENCE_NETWORK / EDGE_TABLE_science-yixue; nearest neighbor E-061 breathing rhythm ↔ Mayer wave/autonomic spectrum, L2/open).


    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.

  • g(m) 残余的单参数闭式拟合:α′ 随 m 掉三倍的残余能否压成 1/m 型公式

    Evidence level: B

    修正因子随测量组数掉落的规律,若能压进一个单参数公式,规范就该从「四格查表」瘦成「一行公式」;压不进,就老实保留查表。

    正文(1000-1500字)

    前序批次(承接 mse-balance-floor-criterion 系列)实测了一件事:当一次判读任务被拆成 m 组独立测量时,高置信水平区间的修正因子 g 明显随 m 掉落——m=2、4、8、16 四档的标尺分别为 0.89、0.58、0.34、0.26,粗看就是「组数翻倍、修正因子掉三成」的节奏。当时的裁决把这一规律做成四格查表:m 落在哪档就查哪个数。查表能用,但它是过渡形态——中间的 m(比如 6 或 12)没有对应值,换一个实验设计就得重新标定,写进操作规范的永远是四行数而不是一条规律。能不能再进一步,把 g(m) 压进一个单参数闭式公式,让查表退役?

    桌面上摆着三种候选形态。第一种是 1/m 型:若 g 与组数严格成反比,α′ 乘 m 应是常数。实测 1.78、2.31、2.74、4.15(m=16 档;按保守口径为 3.68),明显非常数——掉落速度比 1/m 慢,说明分母里除了 m 还有别的项。对 g(m) 做极限展开可知,大 m 端的主项约为 1.66/m:1/m 只描述渐近行为,全程不成立。第二种是 1/根号 m 型:α′ 乘根号 m 为 1.26、1.16、0.97、1.04,四点明显更平,但 m=2 端偏离 1 达两成六,外推到小 m 存疑。第三种是精确 g 本身——它不是近似公式而是精确构造,形态比较的基准就该是它。

    比较的方法是推导加比对:把 1/m、1/根号 m 两种单参数形态与四个实测点做最小二乘拟合,以精确 g 为基准算最大相对偏差。统计文献对这类「闭式近似与精确构造赛跑」有成熟范式:非中心参数与效应量指数的精确区间构造(PMID 36725775)、按小样本量定制区间形状以保覆盖(PMID 20231867)、少研究场景下闭式近似与精确方法的经验比对(PMID 30634920),都指向同一条原则——近似公式的取舍必须用精确构造做基准、用覆盖做裁判,不能只看点拟合好不好看。

    裁决标准分三条。其一,若精确 g 与某个单参数形态在四点的最大相对偏差小于 0.05,该形态进规范,四格查表退役为公式——规范从查表瘦成一行式,中间 m 的插值问题自动消失。其二,若三种形态偏差全部大于 0.10,说明掉落规律里没有干净的单参数骨架,保留四格查表,不硬凑公式。其三,中间情形最微妙:若 1/根号 m 在 m=2 端偏差最大、且方向系统性偏保守(区间偏宽),按「保守方向可接受」降格采用——偏宽只是多花一点精度,偏窄才会漏批次,这个不对称性是统计修正工程化里的标准让步,修正型校正方法同样在精度与保守之间取工程平衡(PMID 37399845)。

    m=16 档的异常也值得一说:α′ 乘 m 在 m=16 端冲到 4.15(保守口径 3.68),与前三档的平滑上升明显脱节——这既可能意味着掉落规律在大 m 端还有二阶以上的修正项,也可能与该档的自由度设置有关。若最小二乘显示精确 g 与单参数形态的偏差集中在 m=16 一格,规范就不该勉强用一条公式覆盖全域,而应写「公式适用于 m 不超过 8,m=16 以上另附一格查表」——公式加尾格的混合形态,比硬塞一个不贴合的公式更诚实。

    公式化还有一个容易被忽略的收益:规范可以反向使用。查表只能事后查——测量做完了才知道落哪档;公式却能在方案评审阶段直接代入预计的组数,提前算出区间半宽会随拆分组数缩多少,从而判断「拆 4 组还是拆 8 组」值不值。也就是说,闭式拟合不只是省下几行表格,它把修正因子从验收工具变成了设计工具。

    这题的本质是可移植性。查表绑死四档 m,公式才绑得住任意设计;公式能不能立,取决于掉落规律的第一主项够不够干净。1.66/m 的渐近主项与根号形态的平坦性相互印证——真实规律大概介于两者之间,靠近哪个端点,正是四点最小二乘要回答的问题。

    证据与局限

    已验证: 所引 4 个 PMID 均真实可查;α′ 随 m 掉落约三倍、四点标尺 0.89/0.58/0.34/0.26 为前序批次仿真实测。
    部分验证: 1/根号 m 形态的平坦性有四点支撑,但 m=2 端外推未经检验;极限展开主项 1.66/m 为推导值,未做逐阶核验。
    待验证: 三种形态的最小二乘比对尚未实际执行;1/根号 m 降格采用的保守代价未量化;二阶修正项是否存在未知。

    相关阅读

    L1 网络状态:该主题尚未进入 L1 对应网络——CORRESPONDENCE_NETWORK 与 EDGE_TABLE_science-yixue 均无「修正因子闭式拟合 / g(m) / 查表退役」对应边;最近邻为 E-060(背俞穴组 与 HRV 分频段的对应边,状态 open,属 L1 升 L2 的可测入口),与本文的方法学主题无直接对应关系。

    证据等级: B
    参考文献: PMID 36725775, PMID 20231867, PMID 30634920, PMID 37399845

    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。

    g(m) residual single-parameter closed-form fitting: can a triple-drop residual in α′ be compressed into an 1/m formula?

    Evidence level: B

    The pattern of the correction factor declining with the number of measurement groups, if compressible into a single-parameter formula, would streamline the standard from a “four-grid lookup table” to a “single-line equation”; if not compressible, the lookup table should be retained as is.

    Body (1,000–1,500 words)

    A preceding batch of trials (continuing the mse-balance-floor-criterion series) tested one specific question: when a single interpretation task is split into m independent measurement groups, the correction factor g of the high-confidence interval drops noticeably as m increases. The scales for m = 2, 4, 8, 16 were 0.89, 0.58, 0.34, and 0.26 respectively; at first glance this follows a rhythm where doubling the number of groups reduces g by roughly one-third. The initial ruling encoded this pattern into a four-row lookup table: whichever row matches the observed m is used. The table works, but it is a transitional form—the intermediate values of m (for example 6 or 12) have no corresponding entry, so changing the experimental design requires re-calibration. What is permanently written into operating procedures remains four rows rather than a single rule. Can we go further and compress g(m) into a single-parameter closed-form formula, allowing the lookup table to retire?

    Three candidate forms lie on the table. The first is of type 1/m: if g is strictly inversely proportional to the number of groups, then α′ multiplied by m should be constant. The measured values are 1.78, 2.31, 2.74, and 4.15 (for m = 16; using the conservative scale this gives 3.68). These are clearly not constant—the drop is slower than 1/m, indicating that besides m there are other terms in the denominator. Expanding g(m) to its limit shows that the leading term at large m is approximately 1.66/m; the 1/m form only describes asymptotic behavior and does not hold across the entire range.

    The second candidate is of type 1/√m: α′ multiplied by √m yields 1.26, 1.16, 0.97, and 1.04. These four points are noticeably flatter than the 1/m case, yet at m = 2 the deviation from 1 reaches 26%, making extrapolation to small m doubtful.

    The third candidate is the exact g itself—it is not an approximate formula but an exact construction, and therefore it should serve as the benchmark against which the other forms are compared.

    The comparison method combines derivation with direct testing: we fit the two single-parameter forms (1/m and 1/√m) to the four measured points using least squares, and compute the maximum relative deviation from the exact g. Statistical literature offers mature paradigms for this kind of “closed-form approximation versus exact construction race”: exact interval construction using non-centrality parameters and effect-size indices (PMID 36725775), tailoring interval shape to small sample sizes to preserve coverage (PMID 20231867), and empirical comparison of closed-form approximations with exact methods under limited research scenarios (PMID 30634920). All point to the same principle: the choice of an approximate formula must be judged against the exact construction and validated by coverage, not merely by how well it fits the observed points.

    The ruling criteria are threefold. First, if the maximum relative deviation between the exact g and a single-parameter form across the four points is less than 0.05, that form enters the specification and the four-row lookup table is retired in favor of a formula—the specification shrinks from four rows to one line, and the interpolation problem for intermediate m disappears automatically.

    Second, if all three forms show deviations greater than 0.10, this indicates that the drop pattern lacks a clean single-parameter backbone; we retain the four-row lookup table and do not force a formula.

    Third, the intermediate case is most subtle: if 1/√m has the largest deviation at m = 2 and that deviation is systematically conservative (the interval is wider), we downgrade to this form under the principle of “conservative direction acceptable.” A wider interval merely costs a little precision; a narrower interval risks missing batches. This asymmetry is the standard concession in statistical correction engineering, and adaptive correction methods likewise balance precision against conservatism (PMID 37399845).

    The anomaly at m = 16 also deserves mention: α′ multiplied by m reaches 4.15 (3.68 on the conservative scale), which clearly breaks the smooth rise of the first three rows. This could mean that the drop pattern still has higher-order correction terms at large m, or it could relate to the degrees-of-freedom setting for that row. If least-squares analysis shows that the deviation between exact g and any single-parameter form concentrates on the m = 16 row, the specification should not force a single formula to cover the entire range. Instead it should state: “the formula applies for m ≤ 8; for m > 8 an additional lookup row is provided.” A hybrid of formula plus tail-row is more honest than hard-fitting a mismatched formula everywhere.

    Formulating the rule also yields an often-overlooked benefit: the specification becomes reversible. A lookup table can only be consulted after measurement—the task is done, then we check which row applies. A formula can be plugged in during the design review stage, using the planned number of groups, to predict how much the interval half-width will shrink as the number of splits increases, thereby deciding whether splitting into 4 or 8 groups is worthwhile. In other words, closed-form fitting does more than save a few rows of tables; it transforms the correction factor from an acceptance tool into a design tool.

    The core issue is portability. The lookup table is locked to four fixed values of m; only a formula can bind arbitrary designs. Whether the formula stands depends on how clean the leading term of the drop pattern is. The asymptotic leading term 1.66/m and the flatness of the √m form corroborate each other—the true pattern likely lies between them, closer to whichever endpoint the least-squares fit indicates.

    Evidence and Limitations

    Verified: All four cited PMIDs are verifiable; α′ drops by approximately threefold with increasing m, and the four-scale markers 0.89/0.58/0.34/0.26 reflect empirical measurements from prior simulation batches.
    Partially Verified: The flatness of the 1/√m form is supported by four points, yet extrapolation beyond m=2 remains untested; the leading term 1.66/m of the asymptotic expansion is derived theoretically and has not undergone step-by-step verification.
    Pending Verification: Least-squares comparison of the three forms has not yet been executed; the conservative cost incurred by downgrading to 1/√m remains unquantified, and the existence of second-order correction terms is unknown.

    Related Reading

    L1 Network Status: This topic has not yet entered the L1 corresponding network—neither CORRESPONDENCE_NETWORK nor EDGE_TABLE_science-yixue contains edges for “correction factor closed-form fitting / g(m) / table retirement”; the nearest neighbor is E-060 (the correspondence edge between Back-Shu acupoint groups and HRV frequency bands, status open, serving as a measurable entry point for L1 elevation to L2), which bears no direct relationship to the methodological theme of this paper.

    Evidence Level: B
    References: PMID 36725775, PMID 20231867, PMID 30634920, PMID 37399845

    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.

  • 95% 档的「反向富余」能不能兑现:按估计的 g 收窄区间,覆盖还守得住吗

    Evidence level: B

    仿真里算出来的红利是对着真分布算的;实践中你手里只有估计值——用 ĝ 收窄区间之前,先问估计噪声会不会把红利吃掉甚至反咬一口。

    正文

    先回顾这笔账从哪来。前面那篇 g 因子文章(95% 档高置信自由度校正的尾部函数)报告过一个反直觉结果:95% 水平下 72 个格子的校正因子 g 全部落在 0.94 到 1.02 之间,其中 w=0.75、ν=3 那一格 g=0.936——意思是修正后的区间比实际需要的还宽了约 6%。这是重尾分布的一个副产品:截尾把尾巴削掉了,但肩膀也变轻了,于是在 95% 这个不算极端的水平上,保守性反而多出了一点。我们管这叫「反向富余」——不是要加宽区间,而是有机会把区间收窄一点、让判读更锐利,覆盖还守得住 0.95。

    问题在于:这个 g<1 是仿真里对着真分布算出来的。实践中没有人知道真分布,g 只能靠标定样本去估计。一旦从「真 g」换成「估计的 ĝ」,故事就完全变了——ĝ 有抽样误差,误差方向不定:这一批数据估出来 0.91,下一批可能估出 0.97。按 ĝ 收窄区间,估小了就把区间切得过窄,覆盖直接掉地板;估大了则红利缩水。红利是否可兑现、兑现几成,从未被测过。

    这正是统计学里一个老问题的翻版:用估计出来的修正量去改进区间,覆盖取决于修正量本身的估计精度。Kenward-Roger 型修正之所以在小样本混合模型里被广泛采用,就是因为它对修正量(自由度与协方差的调整)做了高阶处理,使得「修正本身引入的偏差」不至反噬覆盖;而对纵向数据删失场景的对比研究也显示,协方差估计量的偏差校正方式不同,推断的覆盖表现随之不同。换句话说,「有修正」不等于「修正可白拿」——修正量的估计噪声本身就是新的误差源。另一条平行证据来自宽度与覆盖的权衡:为小样本量身定制的区间方法表明,想把区间压窄,必然要在覆盖上付出代价,关键是代价曲线的形状;少研究个数下的似然法元分析模拟同样显示,窄幅收益越大,覆盖被侵蚀的斜率越陡。

    所以可行的检验设计很直接:在 95% 档对每个格子取 g 的点估计,再叠加估计噪声——ĝ 的抽样方差按标定样本量 10、30、100 三档模拟——然后按 ĝ 收窄区间,重测覆盖与平均窄幅。三档样本量对应三种物理情形:标定样本 10 个时 ĝ 噪声很大,收窄很可能切过头,覆盖掉到 0.93 以下;标定 100 个时噪声小,红利接近全额兑现,平均窄幅接近那 6%;中间档 30 个是分水岭——统计修正文献的经验是,修正量的相对标准误要压到修正幅度的一半以下,修正才净赚不赔。

    还有一层不对称性值得点明:保守(区间偏宽)的代价是判读钝一点,永远不会出错;激进(收窄过头)的代价是假阳性,会直接污染结论库。所以红利兑现的举证责任在收窄方——不是「没证据说不行就收」,而是「有证据说行才收」。这与整个判读规范的设计哲学一致:宁可钝,不可错。

    判读口径事先写死,分三档:若标定样本 30 以上时收窄后覆盖仍不低于 0.945、且平均窄幅至少 3%,则反向富余可写进规范,95% 档按 g 收窄;若收窄后覆盖掉到 0.93 以下,则红利不可兑现,95% 档维持原闭式不动;若窄幅超过 5% 但覆盖贴着 0.945 的地板,说明收益与风险都偏大,收窄只对 ν≥10 的小富余格子开放。

    这事对气功外气实验为什么重要:外气判读依赖谐波幅值与遍数口径的 95% 区间,重尾是常态,截尾是标配。如果反向富余可兑现,同样的标定成本能换来更锐的判读边界——信号与噪音的分界线提前一格;如果不可兑现,硬收窄就是拿覆盖换好看,判出来的「显著」会是假阳性富集。账要算在兑现之前,不能算在发表之后。

    证据与局限

    已验证:宽度与覆盖的权衡关系有直接文献支撑;修正量(自由度、协方差)的估计精度决定推断覆盖,Kenward-Roger 类修正与偏差校正协方差估计均有系统性模拟证据。

    部分验证:g 因子化框架与 72 格 g 值来自本项目内部仿真,非独立复现。

    待验证:ĝ 估计噪声叠加后按 ĝ 收窄的覆盖表现(三档标定样本量、三档判据),为未做的仿真;反向富余在真实脉象/气功数据上的兑现率未知。

    相关阅读

    证据等级: B
    参考文献: PMID 20231867, PMID 30634920, PMID 37399845, PMID 26265765

    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

    L1 网络状态:该主题尚未进入 L1 对应网络(EDGE_TABLE 与 CORRESPONDENCE_NETWORK 中无「反向富余/g 收窄/覆盖兑现」对应边;最近邻 E-060 标签信度,状态 conflict,KB-317:加权系数在「中低一致性+大样本+类别数多」时精度显著下降)。


    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。

    Can the \”reverse surplus\” of the 95% band be realized: can coverage still hold if narrowed by the estimated g interval?

    Evidence level: B

    95% Confidence Interval “Reverse Surplus”: Can It Be Realized? Narrowing the Interval Using Estimated g — Is Coverage Still Maintainable?

    Simulations calculate expected gains against the true distribution; in practice, only estimated values are available. Before narrowing the interval using ĝ, one must first ask whether estimation noise will consume the expected gain—or even reverse it.

    Body

    First, we review the origin of this calculation. The previous article on the g-factor (the tail function with 95% confidence level, high-confidence degrees-of-freedom correction) reported a counterintuitive result: at the 95% level, all 72 grid correction factors g fell between 0.94 and 1.02, with the cell where w=0.75 and ν=3 yielding g=0.936—meaning the corrected interval became approximately 6% wider than actually required. This is a byproduct of heavy-tailed distributions: truncating the tail lightens the shoulders, so at this non-extreme 95% level, conservatism actually increases slightly. We call this “reverse surplus”—not to widen the interval, but to have the opportunity to narrow it slightly, sharpening interpretation while still maintaining coverage above 0.95.

    The problem is: this g<1 was calculated in simulation against the true distribution. In practice, no one knows the true distribution; g can only be estimated via calibration samples. Once we switch from “true g” to “estimated ĝ”, the story changes completely—ĝ has sampling error, with uncertain direction: one batch of data might estimate 0.91, the next 0.97. Narrowing the interval based on ĝ risks cutting it too narrow if underestimated, causing coverage to drop; or too wide if overestimated, wasting potential gain. Whether this gain is realizable and to what extent has never been tested.

    This mirrors a classic statistical problem: using an estimated correction factor to improve interval coverage depends entirely on the precision of that correction estimate itself. Kenward-Roger-type corrections are widely adopted in small-sample mixed models precisely because they apply higher-order adjustments to the correction factor (degrees of freedom and covariance), ensuring that “bias introduced by the correction itself” does not undermine coverage; comparative studies on censored longitudinal data also show that different bias-correction methods for covariance estimators lead to different inferential coverage performance. In other words, “having a correction” does not mean “the correction is free”—estimation noise in the correction factor itself becomes a new error source. A parallel line of evidence comes from the trade-off between width and coverage: interval methods tailored for small samples indicate that narrowing the interval inevitably incurs a cost in coverage, with the shape of the cost curve being critical; likelihood-based meta-analysis simulations similarly show that under fewer studies, the steeper the slope of coverage erosion as narrowness gains increase.

    Thus, the feasible test design is straightforward: at the 95% level, take point estimates of g for each cell, then layer on estimation noise—simulating ĝ’s sampling variance across calibration sample sizes of 10, 30, and 100—then narrow the interval based on ĝ and retest coverage and average narrowness. These three sample sizes correspond to three physical scenarios: with only 10 calibration samples, ĝ noise is large, narrowing likely cuts too much, dropping coverage below 0.93; with 100 samples, noise is minimal, gain approaches full realization, and average narrowness nears that 6%; the middle tier of 30 samples serves as the threshold—statistical correction literature suggests that for a correction to net positive, the relative standard error of the correction factor must be less than half the magnitude of the correction itself.

    One additional asymmetry deserves emphasis: the cost of conservatism (interval too wide) is merely duller interpretation, never wrong; the cost of aggression (narrowing too much) is false positives, which directly pollute the conclusion pool. Therefore, the burden of proof for realizing gain lies with those proposing narrowing—not “narrow unless evidence says no,” but “narrow only if evidence says yes.” This aligns with the overall design philosophy of interpretation norms: better dull than wrong.

    Interpretation criteria are fixed in advance across three tiers: if calibration samples ≥30 and post-narrowing coverage remains ≥0.945 with average narrowness ≥3%, reverse surplus can be incorporated into the norm, and the 95% level uses g-narrowing; if post-narrowing coverage drops below 0.93, the gain is unrealizable, so the 95% level retains its original closed-form; if narrowness exceeds 5% while coverage hovers just above the 0.945 floor, both gain and risk are excessive, so narrowing is opened only for cells with ν≥10 exhibiting slight surplus.

    This matters critically for qigong external qi experiments: external qi interpretation relies on the 95% interval of harmonic amplitude and cycle count, where heavy tails are normal and truncation is standard. If reverse surplus is realizable, the same calibration cost yields sharper interpretation boundaries—the signal-noise threshold advances by one grid; if unrealizable, forced narrowing trades coverage for aesthetics, making declared “significance” a false-positive enrichment. The account must be settled before publication, not after.

    Evidence and Limitations

    Verified: Direct literature supports the trade-off between width and coverage; estimation precision of correction factors (degrees of freedom, covariance) determines inference coverage, with systematic simulation evidence supporting Kenward-Roger-type corrections and bias-corrected covariance estimation.

    Partially Verified: The g-factor framework and the 72-cell g-values originate from internal simulations within this project, not independent replication.

    Pending Verification: Coverage performance after ĝ estimation noise superposition and subsequent narrowing by ĝ (calibrated across three sample size tiers and three decision criteria) remains un-simulated; the realization rate of reverse surplus on real pulse/qigong data is unknown.

    Related Reading

    Evidence level: B
    References: PMID 20231867, PMID 30634920, PMID 37399845, PMID 26265765

    Evidence level: A = direct PubMed support; B = scientific analogy or indirect evidence; C = traditional consensus or preliminary exploration

    L1 network status: This topic has not yet entered the L1 corresponding network (no edges for “reverse surplus / g narrowing / coverage fulfillment” exist in EDGE_TABLE and CORRESPONDENCE_NETWORK; the nearest-neighbor E-060 label has low credibility, status conflict, KB-317: weighted coefficient accuracy decreases significantly under conditions of moderate-to-low consistency plus large sample size plus many categories).


    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.

  • 折减口径的覆盖-宽度解耦检查:省下的宽度有没有偷走 95% 的信誉

    Evidence level: C

    任何把区间「变窄」当成收益来申报的口径,都必须同时申报覆盖率——只报宽度的折减方案,等于允许用置信水平的透支来伪装成方法学进步。

    正文

    前一篇(设计效应标定误差的侵蚀)已经指出:当标定实验估出的组内相关系数 ρ̂ 被代入折减公式时,侵蚀分两档走——第一档是折减收益缩水,第二档更隐蔽,叫「覆盖与宽度解耦」:区间名义上变窄了,但它实际盖住真值的频率(覆盖率)可能已经跌破标称的 95%。这一篇专门裁决第二档:怎么检查折减口径有没有「偷走信誉」。

    先说机制。折减口径的逻辑是:标定实验额外估出了 ρ̂,把这些信息代入设计效应公式(约等于 1 加上会话数减 1 乘 ρ),重复信息不再全额计入方差,区间因此变窄。但这里有个静默假设——把 ρ̂ 当成了真值。ρ̂ 本身有抽样误差:若真实 ρ 比 ρ̂ 大,设计效应被低估,方差该缩的没缩够,区间还是宽的,问题不大;可若真实 ρ 比 ρ̂ 小,折减把方差压得比实际允许的更窄,区间就会系统性过窄,覆盖率跌破 95%。糟糕的是,「省宽度」与「丢覆盖」在报表上长得很像:两者都表现为数字变小。只考核宽度的口径,会奖励那些把 ρ̂ 压到低位、区间好看但实际不可信的做法——这就是「偷工减料被奖励」的完整链条。

    用一个具体数字看透支有多快。假设判读场景里每例被读 3 遍、真实组内相关是 0.3,设计效应约为 1.4,方差按此膨胀。若标定实验因会话数不足,把 ρ̂ 估成了 0.1,设计效应就掉到约 1.2——方差被压掉约 15%,区间等比变窄,报表上这是「漂亮的一轮改进」。但覆盖率的惩罚是非线性的:对正态近似区间,方差被低估 15% 大约意味着真实置信水平从 95% 掉到 91% 附近;低估 30% 就逼近 85%。也就是说,标定越省钱(会话数越少、ρ̂ 越不可信),折减口径的账面收益越大、真实信誉漏得越快,两者是同一枚硬币的两面。这就是为什么检查必须成对申报:单看宽度,你永远看不出硬币已经翻了面。

    这不是我们发明的担忧,而是整类「把参数不确定性计入设计」问题的通用形态。群组随机试验文献里,当只有一个 ICC 估计可用时,用它的置信上限做保守样本量估计是可行但偏低效的路径(PMID 34477455);更新的 assurance / hybrid 方法干脆把参数不确定性直接纳入功效与样本量的计算,而不是事后假装参数已知(PMID 39935065)。信度研究领域同样如此:按预期置信区间宽度规划研究的框架(PMID 11746313)与精度导向的样本量设计(PMID 24338600)都强调,宽度目标只有建立在正确的方差结构上才有意义;而样本量不足时重测信度估计本身会失真(PMID 40119099)。共同教训是:参数有误差时,「名义窄」和「真覆盖」是两回事。

    落到我们的判读场景,检查方案是仿真:沿用前一篇的 ρ̂ 抽样误差网格,在每一档标定会话数 n_s 上同时记录两个量——实际覆盖率(名义 95% 区间真正盖住真值的比例)和平均宽度。把两个量画在同一张图上,得到一条「覆盖-宽度前沿」:每一个口径策略对应前沿上的一个点,要么多覆盖、要么多省宽,不可能兼得。判据由此定死:若 n_s 达到 20 时覆盖率仍不低于 93%,且宽度还保留了已知 ρ 下理论收益的七成以上,折减口径可以在附加「覆盖检查」条款后放行;若覆盖率跌破 90%,说明折减在系统性透支置信水平,必须改用保守校准——例如按 ρ 估计的上限而非点估计来折算设计效应——才允许进入申报规范。

    对读者的意义:当我们宣称「标定让结论更精准」时,这句话只有在覆盖率不缩水的前提下才成立。覆盖-宽度前沿就是防止「精准」这个词被滥用的审计工具——省下的宽度,必须是用真信息换的,不是用信誉换的。

    证据与局限

    已验证: 把 ICC 不确定性计入设计效应的必要性在群组随机试验领域有直接文献支撑(PMID 34477455);assurance 类方法处理参数不确定性的框架已成熟(PMID 39935065);信度研究的宽度导向样本量规划是既成方法学(PMID 11746313、24338600)。

    部分验证: 「覆盖-宽度前沿」作为折减口径的放行判据是本系列提出的设计,其阈值(覆盖 93% 或 90%、收益保留 70%)为工作约定,未经独立文献校准;覆盖率对 ρ̂ 偏差方向的敏感性边界需仿真确定。

    待验证: 折减口径在 ρ̂ 抽样误差网格下的实际覆盖率曲线;保守校准(按上限折算)相比点估计折算的宽度代价;n_s 台阶位置与覆盖安全线的联合取舍(承接下一篇最小样本量台阶定位)。

    相关阅读

    证据等级: C
    参考文献: PMID 34477455, PMID 39935065, PMID 24338600, PMID 11746313, PMID 40119099

    L1 网络状态:该主题尚未进入 L1 对应网络(CORRESPONDENCE_NETWORK.md 与 EDGE_TABLE_science-yixue.md 中无「折减口径 × 覆盖-宽度解耦」对应边);最近邻为 E-061「呼吸节律(0.1 Hz 谱峰)→ Mayer 波 / 自主神经谱」,状态 L2+ / open,原文摘录:「判别规格由『屏气 vs 定速呼吸』升级为『MSNA + HRV 分频段 + 呼吸操控 + 血压』四通道同测……判据尚未被联合实证检验,未升 L3」。本文属测量学方法课题,证据等级按规则封顶 C。

    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。

    Decoupling Coverage-Width Check of Reduced Caliber: Did the Saved Width Steal 95% of Credibility?

    Evidence level: C

    Decoupling Coverage-Width in Reduced Aperture: Has the Saved Width Stolen 95% of Credibility?

    Any aperture that declares narrowing the interval as a gain must simultaneously declare coverage; reporting only reduced width equates to permitting confidence-level overdraft to masquerade as methodological progress.

    Body

    The preceding article (Erosion by Design-Effect Calibration Error) established that when the intraclass correlation coefficient estimate $\hat{\rho}$ derived from calibration experiments is substituted into the design effect reduction formula, erosion proceeds in two tiers: the first tier involves a shrinkage of the reduction benefit, while the second tier is more insidious, termed “decoupling of coverage and width”: although the interval nominally narrows, its actual frequency of covering the true value (coverage) may fall below the nominal 95%. This article specifically adjudicates the second tier: how to verify whether the reduction criterion has “stolen credibility.”

    First, the mechanism. The logic of the reduction criterion is as follows: calibration experiments additionally estimate $\hat{\rho}$; substituting this information into the design effect formula (approximately $1 + (n-1)\hat{\rho}$) means redundant information is no longer fully counted in the variance, and thus the interval narrows. However, there is a silent assumption here—treating $\hat{\rho}$ as the true value. Since $\hat{\rho}$ itself possesses sampling error: if the true $\rho$ is larger than $\hat{\rho}$, the design effect is underestimated, the variance is not shrunk sufficiently, and the interval remains wide; this is not problematic. Conversely, if the true $\rho$ is smaller than $\hat{\rho}$, the reduction compresses the variance narrower than actually permitted, causing the interval to become systematically too narrow, and coverage drops below 95%. The problem is that “saving width” and “losing coverage” look very similar on the report: both manifest as smaller numbers. A criterion that assesses only width rewards approaches that suppress $\hat{\rho}$ to low values, yielding an attractive interval but one that is actually untrustworthy—this constitutes the complete chain of “rewarding cost-cutting.”

    To illustrate how rapidly this overdraw occurs, consider a specific numerical example. Suppose in the interpretation scenario each case is read 3 times, the true intraclass correlation is 0.3, and the design effect is approximately 1.4, with variance inflated accordingly. If the calibration experiment, due to insufficient session count, estimates $\hat{\rho}$ as 0.1, the design effect drops to approximately 1.2—the variance is compressed by about 15%, and the interval scales down proportionally; on the report, this appears as “a beautiful round of improvement.” However, the penalty on coverage is nonlinear: for normal-approximation intervals, underestimating variance by 15% roughly implies the true confidence level falls from 95% to near 91%; underestimating by 30% approaches 85%. In other words, the more cost-saving the calibration (fewer sessions, less trustworthy $\hat{\rho}$), the larger the apparent gain of the reduction criterion and the faster the true credibility leaks; these are two sides of the same coin. This is why verification must be reported in pairs: looking at width alone, you can never see that the coin has flipped.

    This is not a concern we invented; it is the generic form of the broader class of problems involving “incorporating parameter uncertainty into design.” In the literature on cluster randomized trials, when only one ICC estimate is available, using its confidence upper bound for conservative sample size estimation is feasible but suboptimal (PMID 34477455); newer assurance/hybrid methods simply incorporate parameter uncertainty directly into power and sample size calculations, rather than pretending parameters are known post hoc (PMID 39935065). Similarly, in reliability research: frameworks planning studies by expected confidence interval width (PMID 11746313) and precision-oriented sample size designs (PMID 24338600) both emphasize that width targets only make sense if built on the correct variance structure; moreover, when sample size is insufficient, test-retest reliability estimates themselves become distorted (PMID 40119099). The common lesson is: when parameters have error, “nominal narrowness” and “true coverage” are two different things.

    Applied to our interpretation scenario, the verification scheme is simulation-based: using the $\hat{\rho}$ sampling error grid from the preceding article, for each calibration session count $n_s$, we simultaneously record two quantities—actual coverage (the proportion of the nominal 95% interval that truly covers the true value) and average width. Plotting both quantities on the same graph yields a “coverage-width frontier”: each criterion strategy corresponds to a point on this frontier, either offering more coverage or saving more width; it is impossible to achieve both. The criterion is thus fixed: if $n_s$ reaches 20 and coverage remains at least 93%, while width retains more than 70% of the theoretical gain under known $\rho$, the reduction criterion may be approved with an added “coverage check” clause; if coverage drops below 90%, it indicates the reduction is systematically overdrawn on confidence level, and conservative calibration must be adopted—for example, using the upper bound of $\rho$ rather than the point estimate to compute the design effect—before entry into the reporting specification.

    Significance for readers: when we claim “calibration makes conclusions more precise,” this statement holds only if coverage does not shrink. The coverage-width frontier serves as an audit tool to prevent the word “precise” from being abused: any saved width must be exchanged for true information, not for credibility.

    Evidence and Limitations

    Verified: The necessity of incorporating ICC uncertainty into design effects is directly supported by literature in the field of cluster randomized trials (PMID 34477455); assurance-type methods for handling parameter uncertainty have established frameworks (PMID 39935065); and reliability research’s width-oriented sample size planning is an established methodology (PMID 11746313, 24338600).

    Partially Verified: The “coverage-width frontier” as the release criterion for reduced bandwidths is a design proposed in this series; its thresholds (coverage of 93% or 90%, benefit retention of 70%) are working conventions not yet calibrated by independent literature; the sensitivity boundary of coverage to the direction of ρ̂ bias must be determined via simulation.

    Pending Verification: The actual coverage curve of reduced bandwidths under the ρ̂ sampling error grid; the width cost of conservative calibration (converted using upper bounds) versus point-estimate conversion; and the joint trade-off between n_s step positions and the coverage safety line (carrying forward to the next minimum sample size step localization).

    Related Reading

    Evidence Level: C
    References: PMID 34477455, PMID 39935065, PMID 24338600, PMID 11746313, PMID 40119099

    L1 Network Status: This topic has not yet entered the L1 correspondence network (no corresponding edge for “discounting aperture × coverage-width decoupling” in CORRESPONDENCE_NETWORK.md and EDGE_TABLE_science-yixue.md); the nearest neighbor is E-061 “Respiratory Rhythm (0.1 Hz spectral peak) → Mayer waves / autonomic spectrum”, status L2+ / open, with original excerpt: “Discrimination specifications upgraded from ‘breath-hold vs constant-rate breathing’ to four-channel simultaneous measurement of MSNA + HRV frequency bands + respiratory control + blood pressure… the criteria have not yet been jointly empirically validated, so L3 has not been reached.” This article belongs to the measurement methodology topic; per rules, its evidence level is capped at C.

    Evidence Level: A = directly supported by PubMed; 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.

  • 最小标定样本量的台阶定位:se_ρ 衰减曲线的拐点到底在 n_s 等于几

    Evidence level: C

    精度随标定会话数的衰减是”先陡后缓”的,但拐点落在哪个数字上,决定外气研究的折减条款是省钱的台阶还是拍脑袋的猜想。

    正文

    在气功外气这类高成本测量里,”标定会话数”(n_s,用于预校准个体基线变异的重复测量次数)是最直接的成本旋钮。前序分析(df-calibration-error-erosion-discount-gain)预判:标定误差 se_ρ 随 n_s 衰减,先陡后缓,且给出过一个”n_s ≥ 20″的判据。但这个 20 是从哪来的?如果衰减曲线的真正拐点在 10 或 35,那 20 这个台阶就定错了位置——要么花冤枉钱多标了十几次,要么精度根本没到顶、后面的折减口径全建立在沙子上。

    先说机制。精度框架下的信度研究设计,核心不是”显著性检验功效”,而是”置信区间宽度能不能压到目标以内”(PMID 24338600 系统给出了按预期宽度反推样本量的设计框架)。ICC 类系数的标准误随标定会话数近似按 1/√n_s 衰减,但衰减的”坡度”不是均匀的:会话数从 5 加到 15,标准误大约砍掉一半;从 25 加到 35,可能只砍掉 15%。这种边际收益递减正是”台阶”的数学来源——把 se_ρ 曲线对 n_s 求二阶导,符号由负转正的位置就是拐点,也是”再加会话数开始不划算”的转折。

    解析这条路已经有现成的闭式工具。一致性 ICC 的渐近置信区间与样本量公式已给出显式的方差表达式(PMID 39285135),双因素方差分析模型下 ICC 的置信区间与样本量计算也有完整推导(PMID 40405411)。把 n_s 从 5 扫到 50 代进这些闭式,画出 se_ρ 曲线,拐点位置是可以直接算出来的,不需要任何实验。

    但闭式公式假设 ρ 真值已知,而实际标定是拿估计值在跑。模拟研究证实:小样本下 ICC 估计的方差本身会随样本量明显波动,样本量不足时重测信度估计不仅不准、还不稳定(PMID 40119099)。所以第二条路是仿真:在 df-calibration-error-erosion-discount-gain 的误差网格上,对每个 n_s 实测”收益保留率”(相对大样本精度的保留比例),求边际收益曲线的拐点。两条路若给出同一个拐点,台阶才算坐实。

    判据写得死死的:若解析拐点与实测拐点一致且都不超过 20,最小标定样本量按解析台阶取整执行,n_s≥20 判据保留;若实测拐点大于 30,说明原判据定错了位置——此时标定实验的成本必须重估,折减路线与”直接用例数口径”的全成本对比要重算,因为省下的标定钱可能不够补精度损失。另外,若 ρ 的真值本身带较大不确定度,设计时应对单点估计加保守余量(PMID 34477455 在整群试验中示范过”单一 ICC 估计值带不确定度时如何放大样本量”;PMID 41239778 则给出了假设检验口径下 ICC 样本量的另一套算法,可作交叉核对)。

    还要注意一个常被忽略的维度:拐点位置不只由会话数决定,还由”每会话重复次数”决定。同样的精度目标,会话内多重复几次可以换会话数往下压,两条坐标轴之间存在替代关系——只扫 n_s 一条轴,找到的可能是一个”被重复次数补贴过的伪拐点”。稳妥的做法是把(n_s × 每会话重复次数)当二维网格扫,在每一格看 se_ρ,台阶要在二维意义上都平了才算真平。这直接连回覆盖-宽度那条线(df-discount-coverage-width-frontier):折减省下的宽度预算,若被伪拐点偷走精度,覆盖率同样会塌。

    对易学-频率桥接研究,这条看似纯统计的讨论其实卡着脖子:外气实验的单次会话成本高,标定预算定在哪一格,直接决定这类研究在精度上有没有资格进证据分层。台阶定位错了,要么全领域被一个过高的门槛挡在门外,要么一群”精度不够却自称达标”的研究混进来。

    证据与局限

    已验证:精度导向的信度研究设计框架成熟(PMID 24338600);ICC 的闭式方差与样本量公式在多种模型下成立(PMID 39285135、40405411、41239778);样本量不足会同时损害重测信度估计的准确性与稳定性(PMID 40119099)。

    部分验证:se_ρ 衰减”先陡后缓”的形状在精度框架下有理论依据,但”先陡后缓”到什么程度、拐点具体在几,取决于 ρ 真值与每会话重复次数,尚未在本项目的误差网格上实测。

    待验证:拐点是否 ≤20(决定 n_s≥20 判据存废);解析拐点与仿真边际收益拐点是否收敛;拐点 >30 时折减路线与例数口径的全成本重比。均为可执行的解析/仿真任务,结论一旦翻转,须回头修订 df-calibration-error-erosion-discount-gain 的折减条款。

    相关阅读

    证据等级: C
    参考文献: PMID 24338600, PMID 39285135, PMID 40405411, PMID 41239778, PMID 40119099, PMID 34477455

    L1 网络状态:本主题(最小标定样本量台阶 / se_ρ 拐点定位)尚未进入 L1 对应网络;最近邻为 E-060 标签信度前提边(状态 conflict,KB-317/KB-333 记录样本量单位错配与精度区间错配,PMID 40298274、25090041、33853549)与 E-061 的功率约束注记(”0.1 Hz 信号最弱,样本量估算必须吃进去”)。本文为纯测量学方法学课题,证据薄弱,仅作 C 级方法学预判,不构成机制验证。


    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。

    Stepwise Localization of Minimum Calibration Sample Size: Where Does the Inflection Point of the se_ρ Decay Curve Occur at n_s?

    Evidence level: C

    Minimum Calibration Sample Size: Stepwise Positioning of the Inflection Point in the se_ρ Decay Curve

    The decline in precision with increasing calibration sessions follows a “steep then gradual” trajectory; however, the specific numerical value of the inflection point determines whether the reduction clause for external Qi research represents a cost-saving step or an arbitrary conjecture.

    Term Translation
    外气 external Qi (environmental pattern and timing)
    折减条款 reduction clause
    拍脑袋 snap judgment (unsubstantiated assumption)

    Body

    In high-cost measurements such as Qi Gong external Qi, the “number of calibration sessions” (n_s, the number of repeated measurements used to pre-calibrate individual baseline variation) is the most direct cost knob. Prior analysis (df-calibration-error-erosion-discount-gain) predicted that the calibration error se_ρ decays with n_s, initially steeply and then gradually, yielding a criterion of “n_s ≥ 20”. But where does this 20 come from? If the true inflection point of the decay curve lies at 10 or 35, then setting the step at 20 places it incorrectly—either wasting money on over-calibration by ten or more sessions, or failing to reach peak precision, leaving subsequent reductions built on sand.

    First, the mechanism. In a reliability design under the precision framework, the core is not “statistical power of significance testing” but whether “the confidence interval width can be compressed within the target” (PMID 24338600 systematically provides a design framework for back-calculating sample size from expected width). The standard error of ICC-like coefficients decays approximately as 1/√n_s with the number of calibration sessions, but the slope of decay is not uniform: increasing sessions from 5 to 15 cuts the standard error by roughly half, whereas increasing from 25 to 35 may reduce it by only 15%. This diminishing marginal return is precisely the mathematical source of the “step”—taking the second derivative of the se_ρ curve with respect to n_s, the point where the sign shifts from negative to positive marks the inflection, which is also the turning point where “adding more sessions becomes uneconomical.”

    Analytical paths already have closed-form tools. Asymptotic confidence intervals and sample size formulas for consistency ICC provide explicit variance expressions (PMID 39285135), and under two-factor ANOVA models, complete derivations exist for ICC confidence intervals and sample size calculations (PMID 40405411). Plugging n_s values from 5 to 50 into these closed forms and plotting the se_ρ curve allows direct calculation of the inflection point, requiring no experiment.

    However, closed-form formulas assume the true value ρ is known, whereas actual calibration runs on estimated values. Simulation studies confirm that under small samples, the variance of ICC estimates itself fluctuates noticeably with sample size; insufficient samples lead to retest reliability estimates that are not only inaccurate but unstable (PMID 40119099). Thus, the second path is simulation: on the error grid of df-calibration-error-erosion-discount-gain, measure for each n_s the “yield retention rate” (the proportion of large-sample precision retained), and locate the inflection of the marginal yield curve. The step is only firmly established if both paths yield the same inflection point.

    The criterion is strictly defined: if the analytical inflection and the empirical inflection coincide and both do not exceed 20, the minimum calibration sample size follows the analytical step (rounded), and the n_s ≥ 20 criterion is retained. If the empirical inflection exceeds 30, it indicates the original criterion was misplaced—under such conditions, calibration experiment costs must be re-evaluated, and the reduction route versus “direct count-based” full cost must be recalculated, because money saved on calibration may not suffice to compensate for precision loss.

    Furthermore, if the true value of ρ carries substantial uncertainty, the design should add conservative margins to single-point estimates (PMID 34477455 demonstrated in cluster trials how to amplify sample size when a single ICC estimate carries uncertainty; PMID 41239778 provides an alternative algorithm for ICC sample size under hypothesis testing frameworks, useful for cross-validation).

    One often-overlooked dimension must also be addressed: the inflection point depends not only on the number of sessions but also on “repetitions per session.” For the same precision target, increasing repetitions within a session can offset a reduction in the number of sessions; there exists a trade-off between these two axes. Scanning only the n_s axis may yield a “pseudo-inflection point subsidized by repetitions.” A prudent approach is to sweep the two-dimensional grid of (n_s × repetitions per session), evaluating se_ρ at each cell; the step is truly flat only when it levels out in both dimensions. This directly connects to the coverage-width frontier (df-discount-coverage-width-frontier): width budget saved by reduction, if stolen by a pseudo-inflection point, causes coverage to collapse.

    For Yijing Studies–Frequency bridging research, this seemingly purely statistical discussion is critical: the high per-session cost of external Qi experiments means where the calibration budget is placed directly determines whether such studies qualify for evidence tiers based on precision. Misplacing the step either blocks the entire field behind an unnecessarily high threshold or allows in a swarm of studies that claim adequacy despite insufficient precision.

    Evidence and Limitations

    Verified: Reliability research designs oriented toward precision have mature frameworks (PMID 24338600); closed-form variance and sample size formulas for ICC hold across multiple models (PMID 39285135, 40405411, 41239778); insufficient sample size simultaneously undermines the accuracy and stability of retest reliability estimates (PMID 40119099).

    Partially Verified: The “steep then gradual” shape of se_ρ attenuation has theoretical grounding within the precision framework, yet the extent of this “steep then gradual” behavior and the specific location of the inflection point depend on the true value of ρ and the number of repetitions per session, neither of which has been empirically measured on this project’s error grid.

    Pending Verification: Whether the inflection point is ≤20 (determining the validity of the n_s≥20 criterion); whether analytical and simulation marginal gain inflection points converge; and how the full-cost ratio scales when the inflection point exceeds 30. These are executable analytical/simulation tasks; if conclusions reverse, the discounting clauses for df-calibration-error-erosion-discount-gain must be revised accordingly.

    Related Reading

    Evidence level: C
    References: PMID 24338600, PMID 39235135, PMID 40405411, PMID 41239778, PMID 40119099, PMID 34477455

    L1 network status: This topic (minimum calibration sample size step / se_ρ inflection point localization) has not yet entered the corresponding L1 network; the nearest neighbor is the E-060 label credibility premise edge (status conflict, KB-317/KB-333 records sample size unit mismatch and precision interval mismatch, PMID 40298274, 25090041, 33853549) and the power constraint annotation of E-061 (“0.1 Hz signal is weakest, sample size estimation must be consumed”). This article is a pure measurement methodology topic, with weak evidence, serving only as C-level methodological prediction and not constituting mechanism validation.


    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.

  • 基线气感强度测量的重测信度:分层的前提变量自己稳不稳

    Evidence level: C

    如果「基线气感强度」这个自评量自己都不稳,按强弱分层的路线就是在拿噪声切人群。

    上一篇文章提出的分层设计,把「基线气感强度」当成一个既定的分组变量:先让读者自评强弱,再比较强弱两组的会话内相关。这个设计有一个没检查过的前提——那个用来分组的基础测量本身必须稳定。本文专门补这一课:基线气感强度的自评,间隔两周再测一次,还能对上吗。这不是气功研究特有的麻烦,任何用主观自评当分组变量的研究都要先过这一关。

    支持的一面:工具是现成的,而且不少很稳。 现成得气量表里,中文版针刺感觉量表共十二个描述词、五个因子,重测信度做到组内相关系数 0.92,并与麦吉尔疼痛量表的感觉维度显著收敛(PMID 22617434);更早的南安普顿针刺感觉问卷有十七项四分量表的完整验证(PMID 18576920)。放到更广的自评工具里看,常用自评量表如抑郁自评、自我效能量表、一般健康问卷,在两周间隔的重测中组内相关系数全部超过 0.70(PMID 28758322)。还有一项对分层路线有利的推论:大样本因子分析反复发现,多维度内感受自评量表实际只测到一个「一般因子」(PMID 32955947)。结构塌缩听起来像坏消息,但对信度而言反而是好消息——单因子结构的内部一致性更好,自评分组变量跨过 0.7 信度下限的概率更高。如果基线气感强度跟这些量表的行为一样,重测达到 0.7 以上,分层设计就站得住。

    冲突的一面:换一个情境,同一把尺子会失效。 就是那把十七项的南安普顿问卷,德语版在一项实验条件下(六十三名被试)因子结构无法复现,而且痛感与得气感分不开——酸胀类与麻刺类分量表都与疼痛视觉模拟评分显著相关(相关系数 0.574 与 0.496),重测相关也从验证时的水平掉到 0.538 与 0.603(PMID 22189363)。实验情境里被试的注意状态、预期、环境噪声,都会改变自评的内部结构。更进一步,得气自评本身不能当盲法成功与否的检验指标——真针与假针条件下被试的得气反应是混杂的(PMID 40619110)。把这两条映射回我们的问题:基线气感强度在间隔两周的两次自评之间,可能因为情境切换而漂移,而漂移量与真实的个体差异混在一起。若漂移大到相关低于 0.5,强弱分组就近乎随机切分,分层混合模型只会把抽样噪声放大成假分层效应。

    还有一个相关达标也躲不开的检查。 重测相关高,不等于误差小到可以忽略。同样的量表,个体层面的最小可检出改变可以是组层面的十几倍(PMID 28758322 里抑郁自评的个体层门槛是 10.1 分,组层只要 0.86 分)。也就是说,即使分层设计在组间比较意义上成立,也不能拿单次自评给某个人贴「强气感」的标签。报告时应同时给出个体层与组层两个门槛,缺一不可。

    本文的裁决方案:同一批读者间隔至少两周做两次基线气感强度自评,估计重测相关;再用「单次分组」与「两次均值分组」两种口径分别跑分层会话内相关,比较结论是否翻转。判据明确:重测相关达到 0.7 且两种口径结论一致,分层设计成立,单次自评即可分组;重测相关低于 0.5 或结论翻转,必须先立基线测量规范(多日均值或锚定量表),否则分层路线退役。

    证据与局限

    已验证: 现成得气量表有完整的信度验证记录(PMID 22617434、PMID 18576920);常用自评量表两周重测组内相关普遍超过 0.70(PMID 28758322)。
    部分验证: 量表信度跨情境不稳定——同一问卷在实验条件下因子结构无法复现、痛感与得气感无法分离(PMID 22189363);得气自评不能当盲法检验指标(PMID 40619110)。
    待验证: 基线气感强度自评在读者群体中的重测相关本身;单次分组与均值分组两种口径的结论一致性;个体层与组层最小可检出改变的实测值。

    相关阅读

    证据等级: C
    参考文献: PMID 22617434, PMID 18576920, PMID 28758322, PMID 32955947, PMID 22189363, PMID 40619110

    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

    L1 网络状态:E-060 体验通道(得气)边现存对撞结论——「工具现成(ICC=0.92)但跨情境失效(无法分离痛)」:KB-318(置信高)给出 C-MMASS 重测 ICC=0.92 等现成工具;KB-323(conflict)指出 SNSQ 德语版在实验条件下(n=63)因子结构无法复现且无法区分痛与得气,重测降至 r=0.538/0.603(PMID 22189363),且得气不能当盲法检验指标(PMID 40619110)→ 用得气量表前必须在本组样本上重做因子分析并检验能否分离痛。


    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。

    Test-Retest Reliability of Baseline Qi Sensation Intensity Measurement: Is the Stratification Premise Variable Stable?

    Evidence level: C

    If the self-rated measure of “baseline Qi sensation intensity” is itself unstable, stratifying by strength is equivalent to slicing the population with noise.

    The stratified design proposed in the previous article treats “baseline Qi sensation intensity” as a pre-determined grouping variable: readers first self-rate their strength, and then within-session correlations are compared between the strong and weak groups. This design rests on an unverified premise—that the baseline measurement itself must be stable. This article addresses that gap: can the self-rating of baseline Qi sensation intensity be replicated after a two-week interval? This is not a problem unique to Qigong research; any study using subjective self-rating as a grouping variable must first pass this test.

    Supportive Evidence: Tools are available, and many are stable. Among existing scales for eliciting Qi, the Chinese version of the Acupuncture Sensation Scale comprises twelve descriptors and five factors, achieving a test-retest intraclass correlation coefficient (ICC) of 0.92 and showing significant convergence with the sensory dimension of the McGill Pain Questionnaire (PMID 22617434). An earlier Southampton Acupuncture Sensation Questionnaire underwent full validation across seventeen items and four subscales (PMID 18576920). Looking at broader self-rating instruments, common scales such as the Depression Self-Rating Scale, Self-Efficacy Scale, and General Health Questionnaire all yielded ICCs exceeding 0.70 in two-week retests (PMID 28755947). Furthermore, a relevant inference favors the stratification approach: large-sample factor analyses repeatedly reveal that multidimensional interoceptive self-rating scales actually measure only one “general factor” (PMID 32955947). While this structural collapse sounds negative, it is actually positive for reliability—single-factor structures exhibit better internal consistency, increasing the probability that self-rated grouping variables exceed the 0.7 reliability threshold. If baseline Qi sensation intensity behaves like these scales, achieving retest ICCs above 0.7 would validate the stratified design.

    Conflicting Evidence: Changing context can invalidate the same scale. Consider the seventeen-item Southampton questionnaire again; in a German experimental setting with sixty-three participants, the factor structure failed to replicate, and pain sensation could not be distinguished from Qi sensation—the “aching/swelling” and “tingling/pricking” subscales both correlated significantly with pain visual analog scores (correlation coefficients 0.574 and 0.496), while retest correlations dropped from validation levels to 0.538 and 0.603 (PMID 22189363). In such experimental contexts, participant attentional state, expectation, and environmental noise alter the internal structure of self-rating. Moreover, Qi sensation self-rating itself cannot serve as a blind test—responses under real and sham needle conditions are confounded (PMID 40619110). Mapping these findings back to our problem: baseline Qi sensation intensity may drift between two self-ratings separated by two weeks due to context switching, with the magnitude of drift mixing with true individual differences. If the drift reduces correlation below 0.5, stratification becomes nearly random; consequently, hierarchical mixed models would merely amplify sampling noise into spurious stratified effects.

    One final check is unavoidable: meeting the correlation threshold. High retest correlation does not imply negligible error. For the same scale, the minimum detectable change at the individual level can be ten times larger than at the group level (PMID 28755947; for example, the Depression Self-Rating Scale requires an individual threshold of 10.1 points versus a group threshold of only 0.86). Thus, even if the stratified design holds for between-group comparisons, a single self-rating cannot reliably label an individual as “high Qi sensation.” Reporting must include both individual-level and group-level thresholds; omitting either is insufficient.

    Decision Protocol for this Study: The same cohort of readers will complete baseline Qi sensation intensity self-ratings twice, separated by at least two weeks, to estimate the retest correlation. We will then run stratified within-session correlations using both “single-rating” and “mean-of-two-ratings” grouping methods to compare whether conclusions flip. The criteria are explicit: if the retest correlation reaches 0.7 and both grouping methods yield consistent results, the stratified design is validated, and single self-ratings suffice for grouping. If the retest correlation falls below 0.5 or conclusions diverge, we must first establish standardized baseline measurement protocols (e.g., multi-day averages or anchor scales); otherwise, the stratification approach is retired.

    Evidence and Limitations

    Verified: Readiness scales for Qi (a traditional concept of vital energy) have complete reliability validation records (PMID 22617434, PMID 18576920); commonly used self-report scales show intraclass correlations generally exceeding 0.70 upon two-week retest (PMID 28758322).
    Partially Verified: Scale reliability is unstable across contexts—the same questionnaire fails to replicate factor structure under experimental conditions, and pain sensation cannot be separated from Qi sensation (PMID 22189363); self-report of Qi cannot serve as a blind-test indicator (PMID 40617434).
    Pending Verification: The retest correlation of baseline Qi intensity self-assessment within the reader population; consistency between single-group and mean-group approaches; and measured values for the minimum detectable change at both individual and group levels.

    Related Reading

    Evidence Level: C
    References: PMID 22617434, PMID 18576920, PMID 28758322, PMID 32955947, PMID 22189363, PMID 40619110

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

    L1 Network State: E-060 Experience Channel (Deqi) currently presents conflicting conclusions—「tools are available (ICC=0.92) but fail across contexts (cannot separate pain)」: KB-318 (high confidence) provides ready-made tools such as the C-MMASS retest ICC=0.92; KB-323 (conflict) indicates that the SNSQ German version failed to replicate factor structure under experimental conditions (n=63) and could not distinguish pain from Deqi, with retest reliability dropping to r=0.538/0.603 (PMID 22189363), and Deqi cannot serve as a blind-test indicator (PMID 40619110) → before using the Deqi Scale, factor analysis must be re-conducted on this sample group to verify whether pain can be separated.


    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.