Category: 科学易学

  • 气感量表在长期练习者身上是不是顶到了天花板:分布形态、截断比例与「气感变淡」的口径冲突

    Evidence level: C

    「资深者气感更强」可能只是量表在天花板上的读数:期望、身份认同与内标尺漂移,都能让同一把尺子量出相反的假象。

    正文

    丹田呼吸研究里,「资深练习者的气感评分更高」几乎是默认成立的一句话。但这句话有一个从未被检验的测量学前提:气感量表本身在长期练习者身上会不会顶到天花板? 如果会,「更高」就不再是「更强」,而是「尺子到头了」。

    一、两种方向相反的截断机制

    541 的分析指出,资深人群里存在两个方向相反的截断风险。一部分资深者因期望与身份认同,把评分稳定顶在高档——这是典型的天花板效应(ceiling effect):评分压在量表右端、方差被压缩。另一部分资深者的报告恰恰相反:「气感变淡、变细」——不是强度下降,而是分辨变细,0-10 的粗分格不够用了,右端反而空出来。两种机制给出方向相反的分布变形,而现有研究一律只报均值、不报形态,于是「资深者评分更高」到底是真增强、量表饱和还是语义漂移,根本无法判断。

    天花板效应在自评量表里是已知的结构性问题:高分档集中会把方差压到极限,中心趋势与离散度估计都被扭曲,极端分数占比过高时测量模型甚至无法收敛。通用测量学把天花板/地板比例 15%-20% 设为可接受上限——超过这条线,量表对该人群不再有区分能力;偏度 <−1 的显著右偏分布同样提示饱和。

    二、锚定 vignette:把「尺度漂移」与「真实强度差」分开

    深层问题是:不同年限人群使用 0-10 分的「截断点」可能根本不一样。测量学上这叫报告异质性(reporting heterogeneity)——同一句话,不同群体读出的强度位置不同。锚定 vignette 法是分离这种漂移的标准工具:让所有人读同一段标准化体感描述(例如「练习 20 分钟后腹部温热、如蚁行、绵绵不绝」),用同一把尺子给这段描述打分。因为 vignette 对所有人是同一个客观对象,评分的层间差异反映的是「尺子」的差异而非「体感」的差异。

    但工具本身有前提要检验。WHO 十国 5.2 万样本的实证检验发现:vignette 等值性(同一描述被不同群体读成同一水平)在跨文化、跨社会经济群体间常常被违反,而响应一致性(用评 vignette 的尺子评自己)相对稳健(PMID 26335547);年龄与性别匹配还会改变 vignette 评分方式(PMID 25621079)。所以 vignette 法不是拿来就用,两个前提须先验——这构成设计的第二层价值。

    三、纵向随访:反应转移 vs 真增强

    即使横断面形态正常,纵向随访里还有更隐蔽的陷阱:反应转移(response shift)。自评健康研究反复证明,人在状态变化后会换一把内标尺重新评价旧状态——重校准(recalibration,内标尺变化)、重排序(reprioritization,权重变化)、重构念(reconceptualization,概念内容变化),「then-test」设计可定量测出重校准幅度(PMID 28963988)。搬到气感量表上:若资深者一年后评分平台化、却自评「分辨更细」,这恰是内标尺漂移的签名——不是没进步,是进步换了表达方式,而强度评分把它抹平了。此时应改用分辨力类指标(可辨最小变化)替代强度评分。

    四、可执行的测量学设计

    三件事放进一个设计:① 横断面形态——年限四层各 ≥40 人,报气感评分的均值、SD、偏度、峰度、天花板比例,以及分布是否明显右偏或双峰;② 语义锚定——锚定 vignette 法让各层对同一段标准化体感描述打分,分离「尺度漂移」与「真实强度差异」;③ 纵向验证——子样本 ≥30 人随访 12 个月,看同一人评分是随时间单调上升(真增强)还是平台化/回落(饱和或语义漂移)。分析上两点注意:连续变量不要切档(PMID 16217841);趋势检验功效依赖剂量-反应形态、而形态本身须先描述(PMID 35156237)。

    五、判据

    四档判据:若资深层天花板比例 >20% 或分布显著右偏(偏度 <−1),量表在资深人群中已饱和,「评分随年限升高」不可解释为强度增强;若锚定 vignette 得分的层间差异远小于自评实际得分的层间差异,差异主要来自尺度漂移、非体感变化;若纵向随访显示评分平台化而自评「分辨更细」,须改用分辨力类指标替代强度评分;若形态正常、锚定一致且纵向单调,强度评分可用、既往「资深者气感更强」结论可保留。

    这个检验与主效应标定直接联动:#1202 的观测相关上限取决于气感评分自身的信度,而天花板压缩方差 → 信度打折 → 相关打折——如果量表在资深层饱和,「随年限增强」与「气感-力学相关」都可能是同一把尺子到头后的伪象。

    L1 网络状态:CORRESPONDENCE_NETWORK 节点表生理侧有「气感报告 | 结构化自评」(原文「部位/强度/时序/热麻胀沉浮」),但边表止于 E-050(六十甲子纳音↔频率,层级 L1、状态 open,现代证据栏原文「尚无任何可测来源」),grep「天花板 / 饱和 / 语义漂移 / 分布形态 / 偏度 / 量表」零命中;EDGE_TABLE_science-yixue.md 概念计数有气功(48 篇)、丹田(51 篇)、呼吸(189 篇),但无任何以「气感量表分布形态 / 天花板比例 / 锚定 vignette / 反应转移」为源或目标的边。该主题尚未进入 L1 对应网络,已追加至 science-yixue-l1-gaps.md 第 64 条。

    证据与局限

    已验证: 内感受主观报告与客观准确度可解离(PMID 25451381);自评量表天花板效应压缩方差、扭曲中心趋势与离散度估计(通用测量学标准,天花板比例接受线 15%-20%);锚定 vignette 的前提(等值性、响应一致性)在大型健康调查中被实证检验、等值性常被违反(PMID 26335547、PMID 25621079);反应转移(重校准/重排序/重构念)在自评健康纵向数据中被反复检出(PMID 28963988);连续变量切档损失信息(PMID 16217841)。

    部分验证: 内感受注意力量表已有跨语言效度检验范式可借鉴(德语版 PMID 38478969、荷兰语版 PMID 42126131),但「气感」量表在丹田呼吸文献中从未做过分布形态报告;锚定 vignette 法在体感/气感领域的具体应用尚无先例。

    待验证: ① 气感评分在年限四层上的分布形态与天花板比例实测——它是「资深者评分更高」可读性的前提;② 锚定 vignette 的等值性与响应一致性在气感语境下是否成立;③ 12 个月纵向随访中气感评分的平台化/单调趋势,以及「分辨更细」与强度评分的关系。

    相关阅读

    证据等级: C
    参考文献: PMID 25451381, PMID 26335547, PMID 25621079, PMID 28963988, PMID 38478969, PMID 42126131, PMID 16217841, PMID 35156237


    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。

    Does the Qi Sensation Scale Hit a Ceiling for Long-term Practitioners: Distribution Shape, Truncation Proportion, and the Conflict in Defining \”Diminished Qi Sensation\

    Evidence level: C

    The claim that “senior practitioners have stronger qi sensation” may simply reflect the scale hitting its ceiling: expectations, identity认同, and internal scale drift can all cause the same ruler to measure opposite illusions.

    Main Text

    In Dan Tian breathing research, the statement “senior practitioners score higher on qi sensation” is almost taken as given. However, this claim rests on an untested psychometric premise: Does the qi sensation scale itself hit a ceiling in long-term practitioners? If it does, “higher” no longer means “stronger”; it means “the ruler has run out of room.”

    I. Two Opposing Truncation Mechanisms

    Analysis #541 points out two opposing truncation risks among senior populations. Some seniors, driven by expectations and identity认同, consistently push their scores to the high end—a classic ceiling effect: scores compressed at the right end of the scale, variance reduced. Others report the exact opposite: “qi sensation becomes fainter, finer”—not a drop in intensity, but a refinement of discrimination; the coarse 0–10 scale is insufficient, leaving the right end empty. These two mechanisms produce distribution deformations in opposite directions, yet existing studies report only means, ignoring shape. Thus, whether “senior practitioners score higher” reflects true enhancement, scale saturation, or semantic drift, cannot be determined.

    The ceiling effect is a known structural issue in self-report scales: concentration at the high end compresses variance to its limit, distorting central tendency and dispersion estimates; when extreme scores dominate, the measurement model may fail to converge. General psychometrics sets the ceiling/floor proportion at 15–20% as an acceptable upper limit—exceeding this, the scale loses discriminability for that population; significant right skew (skew < −1) also indicates saturation.

    II. Anchored Vignette: Separating “Scale Drift” from “True Intensity Difference”

    The deeper issue is that different age groups may use fundamentally different “truncation points” on the 0–10 scale. Psychometrically, this is called reporting heterogeneity (reporting heterogeneity)—the same statement is perceived at different intensity levels across groups. The anchored vignette method is the standard tool to separate this drift: all participants read the same standardized somatic description (e.g., “After 20 minutes of practice, the abdomen feels warm, like ants moving, continuous and gentle”) and rate it using the same ruler. Because the vignette is an identical objective object for everyone, inter-group differences in scores reflect differences in the “ruler,” not differences in “sensation.”

    However, the tool itself has prerequisites that must be tested. Empirical validation across ten WHO countries with 52,000 samples found that vignette equivalence (the same description being perceived at the same level across cultures and socio-economic groups) is often violated, while response consistency (using the vignette ruler to rate oneself) is relatively robust (PMID 26335547); age and gender matching also alter vignette rating patterns (PMID 25621079). Thus, the vignette method cannot be applied off the shelf; two prerequisites must be established a priori—this constitutes the second layer of design value.

    III. Longitudinal Follow-up: Response Shift vs. True Enhancement

    Even if cross-sectional shape is normal, longitudinal follow-ups contain more subtle traps: response shift. Self-report health research repeatedly shows that people re-evaluate old states using a new internal ruler after their condition changes—recalibration (recalibration, shift in internal scale), reprioritization (reprioritization, weight change), and reconceptualization (reconceptualization, conceptual content change). A “then-test” design can quantify the recalibration magnitude (PMID 28963988). Translated to the qi sensation scale: if seniors plateau in scores after one year but self-report “finer discrimination,” this is the signature of internal scale drift—not lack of progress, but progress expressed differently, with intensity scores flattening it out. In such cases, discriminability metrics (minimum detectable change) should replace intensity scores.

    IV. Executable Psychometric Design

    Three elements must be integrated into one design: ① Cross-sectional shape—at least 40 participants per of four tenure tiers, reporting mean, SD, skewness, kurtosis, ceiling proportion, and whether the distribution is significantly right-skewed or bimodal; ② Semantic anchoring—using the anchored vignette method to have each tier rate the same standardized somatic description, separating “scale drift” from “true intensity difference”; ③ Longitudinal validation—a subsample of ≥30 individuals followed for 12 months to determine whether the same person’s scores rise monotonically over time (true enhancement) or plateau/drop (saturation or semantic drift). Two analytical notes: do not bin continuous variables (PMID 16217841); trend test power depends on dose-response shape, which itself must be described first (PMID 35156237).

    V. Criteria

    Four-tier criteria: If the ceiling proportion in the senior tier exceeds 20% or the distribution is significantly right-skewed (skew < −1), the scale is saturated for seniors, and “scores rising with tenure” cannot be interpreted as intensity enhancement; if inter-tier differences in anchored vignette scores are far smaller than those in self-rated actual scores, the difference stems mainly from scale drift, not somatic change; if longitudinal follow-up shows score plateauing while self-report indicates “finer discrimination,” discriminability metrics must replace intensity scores; if shape is normal, anchoring consistent, and longitudinal trend monotonic, intensity scores are valid, and the prior conclusion “senior practitioners have stronger qi sensation” can be retained.

    This test directly links to main-effect calibration: the observed correlation ceiling in #1202 depends on the reliability of the qi sensation score itself, and ceiling compression reduces variance → lowers reliability → weakens correlation. If the scale is saturated in the senior tier, both “enhancement with tenure” and “qi sensation–mechanics correlation” may be artifacts of the ruler running out of room.

    L1 Network Status: The CORRESPONDENCE_NETWORK node table has “Qi sensation report | structured self-assessment” (original text: “部位/强度/时序/热麻胀沉浮”) on the physiological side, but the edge table stops at E-050 (六十甲子纳音↔频率, level L1, state open, modern evidence column originally states “no measurable source yet”); grep for “ceiling / saturation / semantic drift / distribution shape / skew / scale” yields zero hits; EDGE_TABLE_science-yixue.md concept count includes qigong (48 articles), Dan Tian (51 articles), breathing (189 articles), but no edges sourced or targeted by “qi sensation scale distribution shape / ceiling proportion / anchored vignette / response shift.” This topic has not yet entered the L1 correspondence network and has been added to item 64 of science-yixue-l1-gaps.md.

    Evidence and Limitations

    Verified: Interoceptive subjective reports can dissociate from objective accuracy (PMID 25451381); ceiling effects in self-report scales compress variance and distort central tendency and dispersion estimates (general psychometric standards, acceptable ceiling proportion 15–20%); the anchored vignette prerequisites (equivalence, response consistency) have been empirically tested in large health surveys, with equivalence often violated (PMID 26335547, PMID 25621079); response shift (recalibration/reprioritization/reconceptualization) has been repeatedly detected in longitudinal self-report health data (PMID 28963988); binning continuous variables loses information (PMID 16217841).

    Partially verified: Interoceptive attention scales have cross-lingual validity paradigms that can be borrowed (German version PMID 38478969, Dutch version PMID 42126131), but the “qi sensation” scale has never reported distribution shape in Dan Tian breathing literature; there are no precedents for applying the anchored vignette method specifically to somatic/qi sensation domains.

    To be verified: ① Measured distribution shape and ceiling proportion of qi sensation scores across four tenure tiers—the prerequisite for interpreting “senior practitioners score higher”; ② Whether anchored vignette equivalence and response consistency hold in the qi sensation context; ③ Plateauing or monotonic trends of qi sensation scores over 12 months, and the relationship between “finer discrimination” and intensity scores.

    Related Reading

    Evidence level: C
    References: PMID 25451381, PMID 26335547, PMID 25621079, PMID 28963988, PMID 38478969, PMID 42126131, PMID 16217841, PMID 35156237


    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.

  • 气感评分与 ΔX5 的主效应到底有多大:这个从未被独立报过的数字,卡着整条调节链的命门

    Evidence level: C

    「随年限增强」是调节命题,而调节命题的前提是主效应先成立——如果全样本相关只有 0.1-0.2,讨论「增强」就是在给噪声拟合斜率。

    正文

    「气感评分能预测呼吸力学的变化」——这是丹田呼吸研究里最常被默认成立的一句话。但这句话里藏着一个从未被独立报过的数字:气感评分与 ΔX5(强迫振荡电抗差)之间的主效应相关系数,到底有多大?

    一、为什么这个数字卡着命门

    研究链条的常见写法是「气感评分 → ΔX5,且该关联随修炼年限增强」。这是典型的调节命题:先要有主效应,才有资格谈调节。若全样本的主效应相关只有 0.1-0.2,「随年限增强」讨论的就是噪声的斜率——在噪声上叠加调节项,得到的是对噪声的更精细拟合,不是对机制的更深入解释。把顺序颠倒过来(先证主效应、再谈年限调节)是 #541 给出的反转,但反转落地需要的一个数——主效应量级——在文献里是空的:现有工作要么拿气感评分当「响应」去看组间差异(真/伪呼吸、修炼者/新手),要么拿它当「预测因子」去看主观结局,极少把它与某个力学参数做成一个可报告的相关系数。

    二、两侧的信度都在打折,观测相关的上限被压得很低

    主效应相关低,可能不是「真的没关系」,而是「两边都测不准」。自变量一侧,内感受研究反复证明主观报告与客观准确度是两套东西:心跳知觉任务里,自报信心与实际准确度的相关常常很低(PMID 25451381);连心跳追踪这类结构化客观任务,短期重测信度也只有中等水平(PMID 36564611)。更系统的证据来自维度模型:Garfinkel 等人对心脏与呼吸两轴的内感受做全面比较,发现客观准确度(accuracy)在心脏与呼吸轴之间彼此解离,而主观元认知意识(awareness)却在两轴之间正相关——换句话说,一个人对自身呼吸的「感觉好不好」与「实际测得准不准」根本不在一条轨道上(PMID 28080971)。气感评分本质是主观体验维度(sensibility),ΔX5 是客观力学变化维度(accuracy),两条轨道本就可能低相关。

    因变量一侧同样不稳。FOT 电抗类参数的重复性公认弱于电阻类,便携与常规 IOS 系统的方法比较显示两系统间存在系统偏差(PMID 42409413);儿童 IOS 小气道指数研究也提示电抗类参数的重测区间明显宽于电阻类(PMID 28834381)。自变量与因变量的信度同时打折,观测相关就被双重压缩:观测相关 ≈ 真相关 × 自变量信度 × 因变量信度。

    三、衰减校正:把「测不准」的折扣算清楚

    观测相关被信度压低,这是回归稀释(regression dilution)的标准问题。UK Biobank 用重复测量对 2858 个变量做回归稀释校正,结果显示校正后多数关联明显放大——有些从无意义翻到有意义(PMID 37336529)。这意味着直接报未校正相关会系统性低估;但反过来,若校正前相关接近 0.3、校正后掉到 0.1-0.3,则说明观测关联主要是信度衰减造成的错觉。所以正确做法不是「报一个数」,而是「同时报两变量的信度 + 衰减校正后的相关」。

    四、一份可执行的标定设计

    这个数要从无到有,需要一次纯相关标定:不分层、不干预,只回答「气感与力学参数到底有多大关联」。≥120 名练习年限连续分布的受试者,同次会话完成(a)盲法结构化气感评分(分半或重测信度同时报出)、(b)标准 FOT 测 ΔX5/ΔAX/ΔR5-20(同会话三次均值)、(c)EELV 与潮气量协变量。主分析算气感与三个力学参数的 Pearson/Spearman 及偏相关(控制 EELV、潮气量、年龄、体型),并做衰减校正。注意把连续预测变量切档是已知的坏主意(PMID 16217841),所以直接报连续相关,不做分层。

    判据可以做成四档:若校正后 |r| ≥0.3 且偏相关仍显著,主效应成立、年限调节课题可以继续;若校正后 |r| 落在 0.1-0.3 而校正前接近 0.3,说明观测关联主要是信度衰减造成的错觉、任何调节分析都不可做;若 |r| <0.1 且校正后仍 <0.1,「气感评分预测力学参数」整条假设链作废、须改判为「气感与呼吸力学是两条独立通道」;若未校正 |r| <0.2 但校正后跃升至 >0.4,则问题在测量精度而非构念无关、方向应转为「先做测量改进」。

    五、为什么这个数必须先于一切

    整条丹田呼吸研究线现在处于一个尴尬位置:调节、交互、趋势检验层层堆叠,唯独最底层的主效应量级没人报过。这不只是遗漏,而是结构性的——没有这个数,后面的每个「随年限增强」「特异增量」「趋势显著」都缺少可归因的锚点。先把这个相关标定出来,才知道后续那些精致的统计设计到底是在检验机制,还是在给噪声化妆。

    L1 网络状态:CORRESPONDENCE_NETWORK 节点表生理侧有「气感报告 | 结构化自评」节点(原文「部位/强度/时序/热麻胀沉浮」),但边表止于 E-050(六十甲子纳音↔频率,层级 L1、状态 open,现代证据栏原文「尚无任何可测来源」),无任何以「气感评分 ↔ 呼吸力学参数(FOT 电抗 X5/AX)/ 主效应相关量级 / 衰减校正 / 内感受准确性-意识解离」为源或目标的边;EDGE_TABLE_science-yixue.md 概念计数有气功(48 篇)、丹田(51 篇)、呼吸(189 篇),但 grep「气感↔电抗 / 主效应 / 衰减校正」零命中。该主题尚未进入 L1 对应网络,已追加至 science-yixue-l1-gaps.md 第 63 条。

    证据与局限

    已验证: 内感受主观报告(awareness/sensibility)与客观准确度(accuracy)是两套可解离的维度,且该解离在心脏与呼吸轴上均成立(PMID 25451381、PMID 28080971);心跳追踪类客观任务短期重测信度仅中等(PMID 36564611);FOT 电抗侧重复性弱于电阻侧、便携与常规 IOS 系统间存在系统偏差(PMID 42409413、PMID 28834381);连续预测变量切档损失信息(PMID 16217841)。

    部分验证: 回归稀释校正会显著放大多数关联(UK Biobank 2858 变量实证,PMID 37336529),但该规律具体到「气感 × 呼吸力学」这一对变量上尚无实测;气感评分自身的信度在丹田呼吸文献中从不报告。

    待验证: ① 气感评分与 ΔX5/ΔAX/ΔR5-20 在连续分布样本上的未校正与衰减校正后相关量级;② 偏相关控制 EELV、潮气量、年龄、体型后主效应是否存活;③ 气感评分的分半/重测信度在练习者群体中的实际数值——它决定整个观测相关的上限。

    相关阅读

    证据等级: C
    参考文献: PMID 25451381, PMID 28080971, PMID 36564611, PMID 42409413, PMID 28834381, PMID 37336529, PMID 16217841


    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。

    How Large Are the Main Effects of Qi Sensation Scoring and ΔX5: This Never-Independently-Reported Number Holds the Lifeline of the Entire Modulatory Chain

    Evidence level: C

    “Enhancement with duration” is a moderation proposition, and the premise of any moderation proposition is that the main effect must first be established—if the correlation across the full sample is only 0.1–0.2, discussing “enhancement” amounts to fitting a slope to noise.

    Body

    “The Qi sensation score can predict changes in respiratory mechanics” — this is the most frequently assumed statement within Dantian breathing research. Yet hidden within this sentence lies a number that has never been independently reported: the magnitude of the main effect correlation between Qi sensation scores and ΔX5 (forced oscillation reactance difference). How large is this correlation, truly?

    I. Why This Number Controls the Critical Point

    The common formulation of the research chain reads “Qi sensation score → ΔX5, and this association strengthens with training duration.” This is a classic moderation proposition: one must first establish the main effect before having the qualification to discuss moderation. If the main effect correlation across the full sample is only 0.1–0.2, then discussing “strengthening with duration” merely describes the slope of noise—superimposing a moderation term onto noise yields a more refined fit of the noise, not a deeper explanation of the mechanism. Reversing the order (proving the main effect first, then discussing duration moderation) is the inversion provided in #541, but landing this reversal requires one number—the magnitude of the main effect—which is currently empty in the literature: existing work either treats the Qi sensation score as a “response” to look for between-group differences (true/false breathing, practitioners/novices), or treats it as a “predictor” to look at subjective outcomes, rarely correlating it with any mechanical parameter into a reportable correlation coefficient.

    II. Reliability on Both Sides is Discounted, Compressing the Upper Limit of Observed Correlation

    The low main effect correlation might not mean “they are truly unrelated,” but rather “both sides are measured unreliably.” On the independent variable side, interoception research repeatedly demonstrates that subjective reports and objective accuracy are two separate dimensions: in heartbeat perception tasks, the correlation between self-reported confidence and actual accuracy is often very low (PMID 25451381); even for structured objective tasks like heartbeat tracking, short-term test-retest reliability is only moderate (PMID 36564611). More systematic evidence comes from dimensional models: Garfinkel et al. conducted a comprehensive comparison of interoception across cardiac and respiratory axes, finding that objective accuracy is dissociated between the cardiac and respiratory axes, while subjective metacognitive awareness is positively correlated across both axes—in other words, an individual’s “feeling of how good their breathing is” and “how accurately they actually measure it” are on fundamentally different tracks (PMID 28080971). The Qi sensation score is essentially a subjective experience dimension (sensibility), while ΔX5 is an objective mechanical change dimension (accuracy); these two tracks may inherently have low correlation.

    The dependent variable side is equally unstable. FOT reactance parameters are universally acknowledged to have weaker repeatability than resistance parameters; comparisons between portable and conventional IOS systems reveal systematic biases between the two systems (PMID 42409413); pediatric IOS small airways index studies also indicate that the retest interval for reactance parameters is significantly wider than for resistance parameters (PMID 28834381). With reliability discounted on both the independent and dependent variables, the observed correlation is doubly compressed: observed correlation ≈ true correlation × reliability of independent variable × reliability of dependent variable.

    III. Attenuation Correction: Calculating the Discount for “Unreliable” Measurements

    When observed correlations are suppressed by reliability, this is the standard issue of regression dilution. UK Biobank applied regression dilution correction to 2,858 variables using repeated measurements, showing that after correction most associations were significantly amplified—some shifting from meaningless to meaningful (PMID 37336529). This means directly reporting uncorrected correlations systematically underestimates the true relationship; conversely, if the pre-correction correlation is close to 0.3 but drops to 0.1–0.3 after correction, it indicates the observed association is primarily an illusion caused by reliability attenuation. Therefore, the correct approach is not “reporting one number,” but rather “simultaneously reporting the reliability of both variables plus the attenuation-corrected correlation.”

    IV. An Executable Calibration Design

    This number must be established from nothing, requiring a pure correlation calibration: no stratification, no intervention, simply answering “how large is the actual association between Qi sensation and mechanical parameters.” ≥120 participants with continuous distribution of practice duration, completing in a single session: (a) blind structured Qi sensation scoring (reporting both split-half or test-retest reliability), (b) standard FOT measurement of ΔX5/ΔAX/ΔR5-20 (mean of three measurements within the same session), and (c) EELV and tidal volume covariates. The primary analysis calculates Pearson/Spearman correlations and partial correlations between Qi sensation and the three mechanical parameters (controlling for EELV, tidal volume, age, body habitus), followed by attenuation correction. Note that binning continuous predictors is a known bad idea (PMID 16217841), so we directly report continuous correlations without stratification.

    Criteria can be set at four levels: if the post-correction |r| ≥ 0.3 and partial correlations remain significant, the main effect is established and the duration moderation topic can proceed; if post-correction |r| falls in 0.1–0.3 while pre-correction was close to 0.3, it indicates the observed association is primarily an illusion caused by reliability attenuation, making any moderation analysis impossible; if |r| < 0.1 and remains < 0.1 even after correction, the entire hypothesis chain “Qi sensation score predicts mechanical parameters” is invalidated and must be rejudged as “Qi sensation and respiratory mechanics are two independent channels”; if uncorrected |r| < 0.2 but jumps to > 0.4 after correction, the problem lies in measurement precision rather than construct irrelevance, and the direction should shift to “prioritize measurement improvement.”

    V. Why This Number Must Precede Everything

    The entire Dantian breathing research line is now in an awkward position: layers of moderation, interaction, and trend tests are stacked, yet the most foundational layer—the magnitude of the main effect—has never been reported. This is not merely an omission, but a structural gap: without this number, every subsequent claim of “enhancement with duration,” “specific increment,” or “significant trend” lacks an attributable anchor. Only by first calibrating this correlation can we know whether those sophisticated statistical designs are truly testing mechanisms or merely dressing up noise.

    L1 Network Status: The CORRESPONDENCE_NETWORK node table shows a physiological-side node for “Qi sensation report | structured self-assessment” (original text: “site/intensity/temporal sequence/heat/numbness/swelling/sinking-floating”), but the edge table stops at E-050 (Sixty Jiazi Na Yin ↔ Frequency, level L1, status open, modern evidence column original text: “no measurable source yet”), with no edges sourced or targeted by “Qi sensation score ↔ respiratory mechanical parameters (FOT reactance X5/AX) / main effect correlation magnitude / attenuation correction / interoceptive accuracy-awareness dissociation.” The EDGE_TABLE_science-yixue.md concept count includes Qigong (48 papers), Dantian (51 papers), Breathing (189 papers), but grepping “Qi sensation ↔ reactance / main effect / attenuation correction” yields zero hits. This topic has not yet entered the L1 corresponding network and has been added to item 63 of science-yixue-l1-gaps.md.

    Evidence and Limitations

    Verified: Interoceptive subjective reports (awareness/sensibility) and objective accuracy are two dissociable dimensions, and this dissociation holds across both cardiac and respiratory axes (PMID 25451381, PMID 28080971); short-term test-retest reliability of heartbeat-tracking objective tasks is only moderate (PMID 36564611); FOT reactance parameters have weaker repeatability than resistance parameters, and systematic biases exist between portable and conventional IOS systems (PMID 42409413, PMID 28834381); binning continuous predictors loses information (PMID 16217841).

    Partially Verified: Regression dilution correction significantly amplifies most associations (UK Biobank 2,858 variables empirical evidence, PMID 37336529), but this pattern has not yet been empirically tested specifically for the “Qi sensation × respiratory mechanics” pair; moreover, the reliability of Qi sensation scores itself is never reported in Dantian breathing literature.

    To Be Verified: ① the uncorrected and attenuation-corrected correlation magnitude of Qi sensation scores with ΔX5/ΔAX/ΔR5-20 across continuously distributed samples; ② whether the main effect survives after partial correlation controls for EELV, tidal volume, age, and body habitus; ③ the actual numerical value of split-half/test-retest reliability of Qi sensation scores within the practitioner population—this determines the upper limit of the entire observed correlation.

    Further Reading

    Evidence Level: C
    References: PMID 25451381, PMID 28080971, PMID 36564611, PMID 42409413, PMID 28834381, PMID 37336529, PMID 16217841


    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.

  • 回顾减首次,符号到底是正是负:十二个月复述间隔上三口径嵌套的符号决定性检验

    Evidence level: C

    回顾减首次这个差的符号,决定回顾性研究高估还是低估气感起点;而推动它的两个机制方向相反,所以符号只能实测,不能先验假定。

    一、为什么符号是命门

    同一个练功者被问三次「第一次感觉到气」,给三个日期:日记里首次记下的那天(首次)、此后连续两次以上随访仍报告「有」的那段起点(首次持续)、十二个月末回头被问出的日期(回顾)。把三者相减得到 D1=回顾−首次、D2=首次持续−首次、D3=回顾−首次持续。

    前两个差各有各的含义,但最致命的是 D1 的符号——它直接决定回顾性研究往哪边倒:若 D1 为正(回顾比首次更晚),回顾性研究就系统性高估气感出现时间,把「功夫上身」往后推;若为负(回顾更早),就高估练功见效速度,把期望效应伪装成实证。母文《「第一次感觉到气」有三个日期》已经判定「净方向先验不可判」,本文要把它变成一个可执行的符号决定性检验——因为两个驱动机制,一个把日期往后挪、一个把事件往多算,谁压过谁,文献里从未有人测过。

    二、日期侧:系统性偏晚的证据相当一致

    日期侧的文献方向惊人一致,全部指向「回顾把事件日期挪近现在」,即前向望远镜。最早记忆被儿童一年后复述,日期被系统性后置(PMID 24588518);首次饮酒年龄随复述逐年抬高(PMID 21544259);关节痛患者相隔七年的两次起病日报,绝对差值近四年(PMID 9430817);早期类风湿队列中,起病日落在基线一月内者从 61% 降到 25%(PMID 15338485)。

    最直接的一个量级来自美国全国青年追踪调查(NLSY):5500 多名青少年在 8 年间三次报告首次物质使用年龄,酒精的首次使用年龄在重测中平均后移 1.67 年,且年轻人复述时普遍把 onset 往更晚处报(PMID 23284228)。若气感起点也服从同一条记忆规律,「回顾」系统性晚于「首次」就不是假设,而是默认。

    三、计数侧:回顾多算,且方向相反

    但症状计数侧给出的是相反的力。305 名老人同期填日记与回顾问卷,回顾报告的症状数显著更多(t=304 为 8.48,P<.0001)(PMID 17420523);181 名头痛患儿对比问卷与四周日记,问卷把头痛强度与时长系统性高估(PMID 11168599)。

    多算往前推(回顾里「第一次」比真实更早)、挪日期往后拉(回顾里「第一次」比真实更晚)——两个机制作用在同一条线上,谁压过谁,就是 D1 的符号。这正是母文判定「净方向不可判」的根:不是文献矛盾,是两个独立的偏差源在互相抵消,且没有任何一项研究把它们的净值量出来过。

    四、更麻烦的是:符号可能不是恒定值

    即便做了配对检验,符号也未必是单一常数。儿童记忆定年研究显示,telescoping 的方向随记忆年龄翻转——48 个月之前的记忆被往前挪(前向),48 个月之后的被往后挪(反向望远镜)(PMID 20818575);而提问方式本身就能改变前移偏差的量(PMID 9829101)。这意味着 D1 的符号可能随复述间隔、随人群、甚至随记忆的新旧变化。所以正确的测法不是「测一次符号」,而是「测符号的稳定性」。

    五、怎么测:同队列三口径嵌套 + 依从分层

    测法零额外成本,全部复用既有队列:同一批练功者把三个口径全部采齐,逐人算 D1/D2/D3;用 Wilcoxon 符号秩检验 D1 的中位数是否为 0,并报 bootstrap 中位差的 95% 区间;再按日记漏记天数四分层,看 D1 的符号在低漏记组与高漏记组之间是否翻转——若翻转,说明两个机制的量级可比、净方向只是抽样的产物。

    六、判据三格

    第一格:D1 的中位差 95% 区间排除 0,且绝对值超过一个随访间隔——口径选择实质决定结论,凡跨研究比较气感起点一律无效,必须按口径分组。第二格:区间含 0——两机制确实互相抵消,口径差可当噪声处理,只需在方法里声明。第三格:符号在高漏记组与低漏记组反向——「净方向」这个概念本身不成立,必须按依从分层报两套结论。

    证据与局限

    已验证: 日期侧回溯存在系统性位移且随复述间隔放大(PMID 24588518、21544259、9430817、15338485);前向望远镜有大规模面板实证、onset 年龄重测后移 1.67 年(PMID 23284228);回顾问卷症状数显著多于日记(PMID 17420523);头痛问卷高估强度与时长(PMID 11168599);telescoping 方向随记忆时期翻转(PMID 20818575);提问方式改变偏差量(PMID 9829101)。

    部分验证: 气感起点的口径差量级——相邻领域有数字,本领域为零;D1 的符号——两侧机制方向相反,实测未做。

    待验证: 十二个月复述间隔上 D1 的中位差与符号;符号是否随日记漏记天数分层翻转;D2(首次持续−首次)是否随依从下降放大。

    相关阅读

    L1 网络状态:该主题尚未进入 L1 对应网络(同簇第 2 条,母文 #1185 已登记第 1 条)。CORRESPONDENCE_NETWORK.md 生理/体验侧节点表虽列有「气感报告 | 结构化自评 | 部位/强度/时序/热麻胀沉浮」这一观测项,但边表中不存在以「气感报告」或「事件时间」为源或目标的边;逐词 grep「自报 / 日记 / 依从 / 回顾 / 口径 / 随访 / 首次 / 事件 / 记忆 / 偏差」全部为零。结构性缺失:节点表登记的是「测什么信号」,没有一栏登记「这个信号是怎么被报告上来的」——自报口径(首次 / 首次持续 / 回顾)、依从率、漏记、符号检验全部无处挂载,故这一整类偏差既进不了网络、也不会被网络拦住。证据等级按硬规则封顶为 C。
    证据等级: C(各条方法学证据均有 PubMed 直接支撑,但「气感起点三口径之差的符号」这一命题本身无任何直接研究,且该主题在 L1 对应网络中无对应边;本文是对一个因变量定义偏倚的符号裁决设计,不是已验证机制)
    参考文献: PMID 24588518, PMID 21544259, PMID 9430817, PMID 15338485, PMID 23284228, PMID 17420523, PMID 11168599, PMID 20818575, PMID 9829101


    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。

    Testing the Sign Decisiveness of Three Nested Apertures on the Twelve-Month Repetition Interval: A Retrospective Review of the First Reduction

    Evidence level: C

    The sign of the difference between the retrospective reduction and the first report determines whether retrospective studies systematically overestimate or underestimate the onset of Qi sensation. Because the two mechanisms driving this difference operate in opposite directions, the sign cannot be assumed a priori; it must be determined empirically through measurement.

    I. Why the Sign is the Critical Factor

    When the same practitioner is asked three times about “the first time they felt Qi,” three distinct dates emerge: the date initially recorded in the diary (first report), the starting point of a continuous period where they reported “yes” on two or more consecutive follow-ups (first sustained), and the date recalled twelve months later (retrospective). Subtracting these yields D1 = retrospective − first, D2 = first sustained − first, and D3 = retrospective − first sustained.

    While the first two differences have specific meanings, the sign of D1 is most critical—it directly determines the direction of bias in retrospective studies. If D1 is positive (retrospective date is later than the first report), retrospective studies systematically overestimate the time of Qi sensation onset, pushing the “arrival of the功夫” (Qi sensation event) backward. If D1 is negative (retrospective date is earlier), they overestimate the speed of the practice’s effect, masking expectancy effects as empirical evidence. The parent article, “Three Dates for ‘The First Time I Felt Qi’,” already concluded that the “net direction cannot be determined a priori.” This article aims to transform that conclusion into an executable symbolic determinancy test—because two driving mechanisms exist, one pushing dates later and the other inflating event counts, and literature has never measured which dominates.

    II. The Date Side: Consistent Evidence of Systematic Delay

    The direction of evidence from date-based studies is remarkably consistent, all pointing toward “retrospective shifting events closer to the present,” i.e., forward telescoping. Early memories are systematically delayed when recounted after one year (PMID 24588518); age of first alcohol use increases with each retest (PMID 21544259); in arthritis patients, the absolute difference between two onset reports seven years apart approaches four years (PMID 9433817); in early rheumatoid cohorts, the proportion of onset within one month of baseline drops from 61% to 25% (PMID 15338485).

    The most direct magnitude comes from the National Longitudinal Survey of Youth (NLSY): over 5,500 adolescents reported their first substance use age three times over eight years, with the average age of first alcohol use shifting backward by 1.67 years upon retest, and younger participants consistently reporting later onset (PMID 23284228). If Qi sensation onset follows the same memory law, “retrospective” being systematically later than “first report” is not an assumption but the default expectation.

    III. The Count Side: Retrospective Overcounting, Opposite Direction

    However, symptom-counting studies provide the opposite force. Among 305 elderly participants filling both diaries and retrospective questionnaires, reported symptom counts were significantly higher in the retrospective condition (t=304 = 8.48, P<.0001) (PMID 17420523); among 181 headache children comparing questionnaires to four-week diaries, the questionnaire systematically overestimated both headache intensity and duration (PMID 11168599).

    Overcounting pushes dates forward (retrospective “first” appears earlier than reality), while date shifting pulls them backward (retrospective “first” appears later). These two mechanisms act along the same line; whichever dominates determines the sign of D1. This is precisely why the parent article concluded “net direction cannot be determined”: not because of contradictory evidence, but because two independent bias sources cancel each other out, and no study has ever measured their net value.

    IV. Complication: The Sign May Not Be Constant

    Even with paired testing, the sign may not be a single constant. Children’s memory dating studies show that telescoping direction reverses with memory age—memories before 48 months are shifted forward (forward telescoping), while those after 48 months shift backward (reverse telescoping) (PMID 20818575); moreover, the questioning method itself alters the magnitude of forward-shifting bias (PMID 9829101). This implies D1’s sign may vary with retest interval, population characteristics, or memory recency. Therefore, the correct approach is not “measure the sign once,” but “test the stability of the sign.”

    V. How to Measure: Three Nested Calibers in One Cohort + Adherence Stratification

    The measurement incurs zero additional cost, fully reusing existing cohorts: the same group of practitioners provides all three calibers, calculating D1/D2/D3 individually; apply Wilcoxon signed-rank test to determine if the median of D1 equals zero, reporting bootstrap 95% confidence intervals for the median difference; then stratify by diary omission days into four groups to examine whether D1’s sign flips between low-omission and high-omission strata—if so, it indicates the two mechanisms’ magnitudes are comparable, and net direction is merely a sampling artifact.

    VI. Three-Grid Criteria

    First Grid: The median difference of D1 excludes zero in its 95% interval, and the absolute value exceeds one follow-up interval—caliber selection fundamentally determines conclusions; any cross-study comparison of Qi sensation onset is invalid unless grouped by caliber. Second Grid: The interval includes zero—the two mechanisms indeed cancel each other out; caliber differences can be treated as noise, requiring only methodological declaration. Third Grid: The sign reverses between high-omission and low-omission groups—the concept of “net direction” itself does not hold; two sets of conclusions must be reported stratified by adherence.

    Evidence and Limitations

    Verified: Date-side retrospective recall shows systematic displacement amplified by retest interval (PMID 24588518, 21544259, 9433817, 15338485); forward telescoping has large-scale panel validation with onset age shifting backward 1.67 years (PMID 23284228); retrospective questionnaires report significantly more symptoms than diaries (PMID 17420523); headache questionnaires overestimate intensity and duration (PMID 11168599); telescoping direction reverses with memory period (PMID 20818575); questioning method alters bias magnitude (PMID 9829101).

    Partially Verified: Magnitude of caliber difference for Qi sensation onset—adjacent fields have numerical values, this field has none; sign of D1—the two mechanisms oppose each other, empirical measurement not yet conducted.

    Pending Verification: Median difference and sign of D1 across twelve-month retest intervals; whether D1’s sign flips based on diary omission strata; whether D2 (first sustained − first) amplifies with declining adherence.

    Related Reading

    L1 Network Status: This topic has not yet entered the corresponding L1 network (cluster item #2, parent article #1185 registered item #1). Although the physiological/experience node table in CORRESPONDENCE_NETWORK.md lists “Qi sensation report | structured self-assessment | location/intensity/timing/thermal-numbing-floating-sinking” as an observation item, the edge table lacks any edges with “Qi sensation report” or “event time” as source or target. A word-by-word grep for “self-report / diary / adherence / retrospective / caliber / follow-up / first / event / memory / bias” yields zero results. Structural gap: the node table registers “what signal to measure,” but no column registers “how this signal is reported”—self-report caliber (first / first sustained / retrospective), adherence rate, omission, sign testing—all have nowhere to attach. Thus, this entire class of bias cannot enter the network nor be blocked by it. Evidence level capped at C by hard rules.
    Evidence Level: C (all methodological evidence has direct PubMed support, but the proposition “sign of the difference among three calibers of Qi sensation onset” itself has no direct study, and this topic lacks corresponding edges in the L1 network; this article is a design for adjudicating sign bias of an outcome variable definition, not a verified mechanism)
    References: PMID 24588518, PMID 21544259, PMID 9433817, PMID 15338485, PMID 23284228, PMID 17420523, PMID 11168599, PMID 20818575, PMID 9829101


    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.

  • 「首次持续减首次」这个差,是体感太模糊还是日记漏了那几天:把依从率放进中介模型,看两个来源各占几成

    Evidence level: C

    把「首次持续 − 首次」这个差值当作「一次性模糊体感」的度量之前,必须先把它的两个来路——真实间歇与日记漏记——拆开,否则记录缺陷会被误读成体验属性。

    正文

    研究气感起点时,「首次持续 − 首次」这个差值被 #1185 定义为「一次性模糊体感」的操作化度量:练功者第一次感觉到气感(首次出现),到第一次能稳定持续地感觉到它(首次持续),中间隔了多少天。这个差值越大,被解读为「气感本身时有时无、不是一出现就站稳」的证据越强。

    但问题是:这个差至少有两条完全不同的来路。第一条是真实间歇——气感确实是间歇性出现的,今天有、明天没有、后天又有,于是「连续两次都感觉到」这一达标规则把它整体推迟。第二条是日记漏记——那几天其实有气感,但练功者忘了记,在日记上留下一个假空白,造成「中间断档」的假象。这两条路的含义南辕北辙:如果是真实间歇,说明气感是非稳态现象,研究就得改用「首次达标强度」而非「首次出现」作为起点口径;如果是漏记,那只要把日记提醒机制做好就能修好,判据口径根本不用动。

    而依从性本身不是随机丢失的,它有明确的结构。九周 EMA 研究显示依从性系统性下降:早间评估每周掉 2%、随机评估每周掉 1%(PMID 37347538);第一天的首次提示应答最差,几乎是个普遍规律(PMID 35537542);早期依从能强预测全程依从——开头守不住的人,后面大概率一直守不住(PMID 34541425)。也就是说,「哪天没记」不是均匀散布的空白,而是集中在研究早期、集中在特定人群里。这个结构意味着:如果不对漏记做处理,D2 差值会被系统性污染——漏记多的那一周看起来就像「气感断档」,其实只是「记录断档」。

    这正是本文要拆的问题。拆法并不复杂:在日记里同时加记两栏——「今天有没有气感」与「今天有没有按时记」。然后对每一个「首次持续 − 首次」的间隔逐日标注:标为「有记录且报告无」的日子计为真实间歇,标为「无记录」的日子计为漏记。这样每个间隔内部的两类天数都有了独立计数,再以 D2 为因变量、真实间歇天数与漏记天数为两个并行中介,做多重中介分解,报各自的比例中介。这是测量学上把「观测缺失」与「真实状态」分离的标准思路——EMA 文献里漏记事件的研究(PMID 33052123)与缺失数据模式分析(PMID 34464327)已经证明,把「没发生」和「没记录」混在一起会把偏差系统性引入估计。

    如果两个来源纠缠到无法分离——真实间歇与漏记相互抑制、总效应小于两条路径之和——那就要升级方法:改用带删失的隐马尔可夫模型(HMM),把「真实气感状态」当隐藏状态、把「有没有记」当观测过程,同时估计两者。这在方法学上是成熟的:HMM 处理缺失响应与间歇性缺失模式(intermittent pattern)已有完整框架(PMID 37035989),非等距纵向记录上的隐多状态建模也有成熟的贝叶斯解法(PMID 32162300)。也就是说,拆不开不是死路,只是意味着必须把「观测过程」和「真实过程」一起建模。

    为什么要这么较真?因为 D2 是两个研究结论的分叉口:把它判成「体验属性」还是「记录缺陷」,直接决定整个气感起点研究是改口径还是改提醒。依从性结构告诉我们漏记不是随机的,中介分解能给出两类来源的占比,HMM 能在最坏情况下兜底。三者加在一起,才够把「首次持续 − 首次」这个差从「一个说不清的间隔」变成「一个有归属的测量量」。

    证据与局限

    已验证: EMA 依从性存在系统性下降结构(早间评估每周降 2%、随机评估每周降 1%,PMID 37347538);首日首次提示应答最差(PMID 35537542);早期依从强预测全程依从(PMID 34541425);事件型 EMA 中「漏记事件」是独立于真实事件的测量问题(PMID 33052123);HMM 可同时处理缺失响应与间歇性缺失模式(PMID 37035989)。

    部分验证: 非等距纵向记录的隐多状态贝叶斯建模(PMID 32162300)证明了不规则观察下分离隐藏状态的可行性;「漏记天数可按依从结构分层预测」的假设有依从性文献支撑,但尚未在气感日记场景实测。

    待验证: ① 真实间歇天数与漏记天数作为并行中介的比例分配——>50% 漏记主导则改提醒机制即可,>50% 真实间歇主导则气感确为非稳态、「首次持续」为唯一口径;② 两中介路径是否相互抑制(总效应 < 两路径之和)——是则须升级为带删失 HMM;③ 依从分层(低/高漏记组)下 D2 的符号是否翻转——翻转则「净方向」是依从的函数,跨研究比较须按依从分层。

    相关阅读

    证据等级: C
    参考文献: PMID 37347538, PMID 35537542, PMID 34541425, PMID 33052123, PMID 34464327, PMID 37035989, PMID 32162300

    L1 网络状态: 「气感报告」↔「事件时间(起病/首次发生)」边缺失的归因层深化(「自报过程/事件时间」簇第 3 条):CORRESPONDENCE_NETWORK 生理/体验侧节点表仅列有「气感报告|结构化自评|部位/强度/时序/热麻胀沉浮」这一观测项(全网络唯一「气感」命中,1 处),边表中不存在以「气感报告」或「事件时间」为源或目标的边;逐词 grep「依从 / 日记 / 漏记 / 真实间歇 / 记录缺陷 / 中介 / 隐马尔可夫」全部为零。该主题尚未进入 L1 对应网络。


    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。

    Is the gap between \”first sustained reduction\” and \”first ever\” due to vague perception or missing diary days: putting adherence into a mediation model to see the proportion con…

    Evidence level: C

    Before treating the difference between “first sustained” and “first” as a metric of “one-time fuzzy perception,” it is necessary to disentangle its two origins—true intermissions and diary omissions. Otherwise, recording defects may be misinterpreted as experiential attributes.

    Main Text

    When investigating the onset of Qi sensation, the difference “first sustained minus first” is operationalized as a measure of “one-time fuzzy perception” (a transient, non-stable sensory experience) #1185. This metric captures the number of days between the first time a practitioner feels Qi (first occurrence) and the first time they can stably and continuously feel it (first sustained). A larger difference is interpreted as stronger evidence that “the Qi sensation itself comes and goes, not stabilizing immediately upon appearance.”

    However, this gap has at least two distinct origins. The first is true intermission—the Qi sensation genuinely appears intermittently (present today, absent tomorrow, present again the next day), causing the “continuous two-day” criterion to delay the overall count. The second is diary omission—the practitioner actually felt Qi on those days but forgot to record it, leaving a false blank in the diary and creating an illusion of “interruption.” These two paths have opposite implications: if it is true intermission, Qi sensation is a non-stationary phenomenon, and the study must switch from “first occurrence” to “first qualifying intensity” as the onset criterion; if it is omission, simply improving the diary reminder mechanism would fix it without altering the judgment criterion.

    Adherence itself is not randomly lost; it has a clear structure. A nine-week EMA study showed systematic decline in adherence: morning assessments dropped 2% per week, random assessments dropped 1% per week (PMID 37347538); the first day’s initial prompt response was worst, nearly a universal rule (PMID 35537542); early adherence strongly predicted overall adherence—those who couldn’t stick at the start likely couldn’t stick later (PMID 34541425). Thus, “which day was missed” is not uniformly distributed but concentrated in the early study phase and among specific subgroups. This structure means that without addressing omissions, the D2 difference will be systematically contaminated—the week with many omissions may appear as “Qi sensation interruption,” though it is actually just “recording interruption.”

    This is precisely the problem this paper addresses. The solution is straightforward: add two columns to the diary—”Did you feel Qi today?” and “Did you record on time today?” Then, for each interval of “first sustained minus first,” mark each day: days marked “recorded but reported no” count as true intermissions, days marked “no record” count as omissions. This gives independent counts of both types within each interval, allowing a multiple mediation decomposition with D2 as the dependent variable and true intermission days and omission days as two parallel mediators, reporting their respective proportion mediated. This is the standard measurement-theoretic approach to separating “observed missingness” from “true state”—EMA literature on omitted events (PMID 33052123) and missing data pattern analysis (PMID 34464327) already proves that mixing “did not happen” with “was not recorded” systematically introduces bias into estimates.

    If the two sources become entangled and inseparable—true intermission and omission mutually suppressing each other, resulting in a total effect smaller than the sum of both paths—then the method must be upgraded: use a hidden Markov model (HMM) with censoring, treating “true Qi state” as the hidden state and “whether recorded” as the observation process, estimating both simultaneously. This is methodologically mature: HMMs have a complete framework for handling missing responses and intermittent missing patterns (PMID 37035989), and Bayesian solutions exist for hidden multi-state modeling on non-equidistant longitudinal records (PMID 32164300). Thus, being unable to separate them is not a dead end; it simply means both the “observation process” and the “true process” must be modeled together.

    Why go to such lengths? Because D2 is the fork where two research conclusions diverge: classifying it as “experiential attribute” or “recording defect” determines whether the Qi onset study must change its criterion or improve reminders. The adherence structure tells us omissions are not random; mediation decomposition yields the proportion of each source, and HMM provides a safety net under worst-case scenarios. Together, these three elements transform “first sustained minus first” from “an ambiguous interval” into “a well-anchored measurement quantity.”

    Evidence and Limitations

    Verified: EMA adherence exhibits systematic decline structure (morning assessments drop 2%/week, random assessments drop 1%/week, PMID 37347538); first-day initial prompt response is worst (PMID 35537542); early adherence strongly predicts overall adherence (PMID 34541425); in event-based EMA, “omission events” are measurement issues independent of true events (PMID 33052123); HMM can simultaneously handle missing responses and intermittent missing patterns (PMID 37035989).

    Partially Verified: Bayesian hidden multi-state modeling for non-equidistant longitudinal records (PMID 32164300) demonstrates feasibility of separating hidden states under irregular observation; the hypothesis that “omission days can be stratified and predicted based on adherence structure” is supported by adherence literature but remains untested in Qi diary scenarios.

    To Be Verified: ① Proportion allocation of true intermission days and omission days as parallel mediators—if >50% omissions dominate, improve reminder mechanism; if >50% true intermission dominates, Qi sensation is confirmed non-stationary and “first sustained” becomes the sole criterion; ② Whether the two mediation paths mutually suppress each other (total effect < sum of both paths)—if so, upgrade to censored HMM; ③ Whether the sign of D2 flips under adherence stratification (low/high omission groups)—if flipped, “net direction” becomes a function of adherence, and cross-study comparisons must be stratified by adherence.

    Related Reading

    Evidence Level: C
    References: PMID 37347538, PMID 35537542, PMID 34541425, PMID 33052123, PMID 34464327, PMID 37035989, PMID 32164300

    L1 Network Status: “Qi sensation report” ↔ “Event time (onset/first occurrence)” edge attribution layer deepened (“self-reported process/event time” cluster item 3): CORRESPONDENCE_NETWORK physiological/experiential node table lists only “Qi sensation report | structured self-assessment | location/intensity/timing/heat-numbness-swell-float-sink” as the observation item (the sole “Qi sensation” hit in the entire network, 1 location); no edges exist with “Qi sensation report” or “event time” as source or target; grepping all terms—”adherence / diary / omission / true intermission / recording defect / mediation / hidden Markov”—yields zero. This topic has not yet entered the corresponding L1 network.


    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.

  • 越补救越糟:两阶段加密窗在粗扫偏了的时候,反而比均匀铺点更差

    Evidence level: C

    先粗扫定位、再在候选区加密,这个补救方案默认了「粗估的点足够准」——而变点估计偏偏在最需要加密的边界处最不可靠,漏窗时加密采到的全是同侧数据,结果比均匀铺点更差。

    正文

    测量「功夫上身」的时间,本质是在估计一条随时间变化的曲线上的拐点:练功多少个月,胸腹相位角才开始出现可检测的位移。上一轮(#1184)我们已经知道,这个拐点的最小可检测位移不是常数,而是拐点位置的函数——拐点落在随访区间两端时,检测能力系统性变差。面对这个困境,一个看起来很聪明的补救方案是两阶段设计:先用几个稀疏的点粗扫一遍,大概定位拐点在哪个区间,再回到那个区间加密采样,把点数花在刀刃上。

    这个方案的底气来自一条统计学的经典结论:对本质非线性的函数,在曲率最大处加密采样,参数恢复最好(PMID 26609742)。问题在于,拐点恰恰是我们要估的东西——曲率最大处在哪,事先并不知道。于是两阶段设计必须依赖一个隐含前提:粗估的点足够准,加密窗能圈住真拐点。

    这个前提恰恰是最不可靠的。变点估计有一个已知的病理:当真拐点靠近边界、或样本量小时,点估计与区间估计会一起失效(PMID 40585262)。也就是说,粗扫阶段最容易出错的地方,正是边界附近——而边界附近又恰是最需要加密的那一段。两件事叠在一起,构成了一个「越补救越糟」的结构:

    如果粗估把真拐点漏在了加密窗外,那么加密采样采到的全部是真拐点同一侧的数据。对这样一个单侧数据堆做拟合,拐点会被牢牢钉在加密窗的边界上——不是因为它真的在那,而是因为窗外没有任何数据能把它拉回来。此时两阶段设计的误差,会比同样点数均匀铺开还要大。

    这不是理论上的杞人忧天。测量频率与随访设计的研究早就提醒过:增加测量点数带来的收益是递减的(PMID 33644405),所以两阶段设计必须胜过「同样点数的均匀铺点」才有意义——否则你就是用更复杂的方案、花同样的钱、拿到更差的结果。而自适应设计的统计文献也反复强调一个核心教训:设计规则里嵌进的估计(比如用中期数据选剂量、选亚组),会把自己的误差带进最终估计(PMID 26459506);变点模型驱动的自适应设计在药理学里早已被检验过,它的可操作特性(operating characteristics)必须靠仿真逐场景标定(PMID 27221807),没有任何理由相信「粗估 → 加密」在气功纵向数据上会自动成立。

    那么怎么裁决?这是一个纯仿真就能回答的问题,不需要额外招募一个人。用 #1184 同一套实测噪声底,把三种设计并排摆开:

    1. 现行 5 点均匀(0/1/3/6/12 月);
    2. 均匀 9 点(与两阶段总点数对齐);
    3. 两阶段:粗扫 3 点等距,按粗估拐点 ±1 个间隔设加密窗、窗内再取 4 点。

    然后让真拐点位置在 1-12 月之间扫描、幅度在 1-10 度之间扫描,各重复 ≥500 次,记录三件事:拐点区间半宽、点估计偏差、以及漏窗率(真拐点落在加密窗之外的比例)。

    判据分三档:

    • 若漏窗率 >20%,且 6-12 月段区间半宽反而比均匀 9 点更差 → 「加密」这条补救路线在该设计下作废,改均匀铺点或连续记录;
    • 若漏窗率 <5%,且半宽优于均匀 9 点 → 两阶段可用,但加密窗必须按粗估的区间上界宽度(而不是点估计)来定,给粗估的误差留出余量;
    • 若两阶段只在拐点落于窗内(2-8 月)时优于均匀、在两端劣于均匀 → 写成条件处方:先做均匀粗扫判定「是否值得进两阶段」,而不是一开始就跳进两阶段。

    这一判的核心不是「两阶段好不好」,而是「补救方案的有效性本身是有条件的」——任何把粗扫当作免费午餐的设计,都必须先证明粗扫的误差不会传染给加密阶段。对研究「练功多久见功」这类问题,这个裁决直接决定后续所有随访协议:是花两倍点数赌一个可能漏窗的两阶段,还是老老实实均匀铺点。

    证据与局限

    已验证: 变点估计在真拐点靠近边界或样本量小时点估计与区间估计共同失效(PMID 40585262);在曲率最大处加密采样对本质非线性函数参数恢复最好(PMID 26609742);测量频率与随访时长对非线性趋势项估计的改善存在收益递减(PMID 33644405);自适应设计中的选择规则会把估计误差带进最终估计(PMID 26459506);变点模型自适应设计的操作特性须仿真逐场景标定(PMID 27221807)。

    部分验证: 两阶段(先粗扫后加密)设计在剂量反应试验中已有损失函数框架支撑(PMID 12036265),但其收益依赖中期估计的准确性,该依赖在纵向随访场景下的表现尚无直接证据。

    待验证: ① 以 #1184 同一套 φ 噪声底代入,两阶段设计的漏窗率在 6-12 月段是否真的 >20%;② 加密窗按「区间上界宽度」而非点估计来定,能否把漏窗率压到 5% 以下;③ 两阶段是否只在拐点落于窗内时优于均匀铺点(条件处方的成立范围)。

    相关阅读

    证据等级: C
    参考文献: PMID 26609742, PMID 40585262, PMID 33644405, PMID 26459506, PMID 27221807, PMID 12036265

    L1 网络状态: 该主题(变点估计的两阶段采样设计是否受粗扫误差污染)尚未进入 L1 对应网络——CORRESPONDENCE_NETWORK 节点表生理侧仅列「呼吸率 / 呼吸模式 | 可穿戴 | 与节律耦合最直接」,无「随访采样设计 / 变点估计 / 加密窗 / 粗扫误差」类测量学节点或边;EDGE_TABLE_science-yixue.md 反抽概念表无「变点 / 采样设计 / 加密窗」条目。该主题已登记至 science-yixue-l1-gaps.md。


    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。

    The more you patch, the worse it gets: Two-stage encryption windows perform worse than uniform point placement when coarse scanning is misaligned.

    Evidence level: C

    The proposed remedy of a coarse scan followed by encryption assumes that “the rough estimate is accurate enough.” Yet, change-point estimation is precisely least reliable at the boundaries where encryption is most needed. When the window leaks, the encrypted sample captures data only from one side of the true turning point. Fitting such unilateral data forces the estimated turning point to the window boundary—not because it is there, but because no data outside the window can pull it back. Consequently, this approach yields worse results than uniform spacing.

    Body

    Estimating the time when “cultivation takes effect” fundamentally involves estimating the inflection point on a curve that changes over time: after how many months does the thoraco-abdominal phase angle begin to show a detectable shift. As established in the previous round (#1184), the minimum detectable displacement at this inflection point is not constant; it is a function of the inflection point’s location itself—detection capability systematically degrades when the inflection point lies near the boundaries of the follow-up interval.

    Faced with this dilemma, a seemingly clever remedy is the two-stage design: first, use a few sparse points for a rough scan to roughly locate which interval the inflection point falls into, then return to that interval for dense sampling, spending the points where they matter most.

    The confidence in this approach stems from a classic statistical conclusion: for inherently nonlinear functions, dense sampling at the point of maximum curvature yields the best parameter recovery (PMID 26609742). The problem is that the inflection point is precisely what we are trying to estimate—the location of maximum curvature is unknown beforehand. Thus, the two-stage design must rely on an implicit assumption: the rough estimate is accurate enough that the encryption window will capture the true inflection point.

    This assumption is precisely the most unreliable part. Change-point estimation has a known pathology: when the true inflection point is near the boundary or the sample size is small, both the point estimate and the interval estimate fail simultaneously (PMID 40585262). In other words, the coarse scan stage is most prone to error exactly at the boundaries—the very region that needs encryption the most. These two factors combine to create a “the more you try to fix it, the worse it gets” structure:

    If the rough estimate misses the true inflection point outside the encryption window, then all data collected during the dense sampling phase comes from only one side of the true turning point. Fitting such unilateral data forces the estimated inflection point to the edge of the encryption window—not because it is truly there, but because no data outside the window can pull it back. Under these conditions, the error of the two-stage design exceeds that of uniform spacing with the same number of points.

    This is not theoretical worry. Research on measurement frequency and follow-up design has long warned that the benefit of increasing the number of measurements diminishes (PMID 33644405). Therefore, the two-stage design must outperform “uniform spacing with the same number of points” to be meaningful—otherwise, you are using a more complex scheme, spending the same money, and getting worse results. Statistical literature on adaptive design also repeatedly emphasizes a core lesson: estimation rules embedded in the design (e.g., selecting dose or subgroups using interim data) carry their own error into the final estimate (PMID 26459506); adaptive designs driven by change-point models in pharmacology have long been tested, and their operating characteristics must be calibrated via simulation across scenarios (PMID 27221807). There is no reason to believe that “rough estimate → encryption” would automatically hold for Qigong longitudinal data.

    So how do we decide? This is a question answerable entirely by simulation, requiring no additional recruitment of a single person. Using the same measured noise floor from #1184, we compare three designs side by side:

    1. Current 5-point uniform (0/1/3/6/12 months);
    2. Uniform 9-point (aligned with the total number of points in the two-stage design);
    3. Two-stage: coarse scan with 3 equally spaced points, setting the encryption window at ±1 interval around the rough estimate of the inflection point, then taking 4 additional points within the window.

    We then scan the true inflection point location across 1–12 months and its amplitude across 1–10 degrees, repeating each scenario ≥500 times, recording three metrics: half-width of the inflection point interval, bias of the point estimate, and window leak rate (the proportion of true inflection points falling outside the encryption window).

    The criteria are tiered:

    • If the window leak rate >20%, and the half-width of the 6–12 month interval is worse than uniform 9-point → the “encryption” remedy route is invalid under this design; switch to uniform spacing or continuous recording;
    • If the window leak rate <5%, and the half-width is better than uniform 9-point → the two-stage design is viable, but the encryption window must be set based on the interval upper-bound width (not the point estimate) of the rough estimate, leaving room for the error of the rough scan;
    • If the two-stage design is superior only when the inflection point falls inside the window (2–8 months), but worse at both ends → formulate a conditional prescription: first perform uniform coarse scanning to determine “whether it is worth entering the two-stage design,” rather than jumping straight into the two-stage design from the start.

    The core of this judgment is not “is the two-stage design good,” but rather “the validity of the remedy itself is conditional”—any design that treats the coarse scan as a free lunch must first prove that the error of the coarse scan does not propagate into the encryption stage. For research questions like “how long until cultivation takes effect,” this verdict directly determines all subsequent follow-up protocols: whether to gamble twice the number of points on a two-stage design that might leak the window, or to stick with uniform spacing.

    Evidence and Limitations

    Verified: Change-point estimation fails simultaneously for both point and interval estimates when the true inflection point is near the boundary or sample size is small (PMID 40585262); dense sampling at maximum curvature optimally recovers parameters of inherently nonlinear functions (PMID 26609742); increasing measurement frequency yields diminishing returns for estimating nonlinear trend terms (PMID 33644405); selection rules in adaptive designs embed their own estimation error into the final estimate (PMID 26459506); operating characteristics of change-point model-driven adaptive designs must be calibrated via scenario-specific simulation (PMID 27221807).

    Partially Verified: The two-stage (coarse scan then encryption) design has a loss function framework supporting it in dose-response trials (PMID 12036265), but its benefit depends on the accuracy of interim estimates, for which there is no direct evidence in longitudinal follow-up scenarios.

    To Be Verified: ① With the same φ noise floor from #1184, does the two-stage design’s window leak rate truly exceed 20% in the 6–12 month interval? ② If the encryption window is set by “interval upper-bound width” rather than point estimate, can the leak rate be suppressed below 5%? ③ Does the two-stage design outperform uniform spacing only when the inflection point falls within the window (the range of validity for the conditional prescription)?

    Further Reading

    Evidence Level: C
    References: PMID 26609742, PMID 40585262, PMID 33644405, PMID 26459506, PMID 27221807, PMID 12036265

    L1 Network Status: This topic (whether the two-stage sampling design for change-point estimation is contaminated by coarse scan error) has not yet entered the L1 corresponding network—the CORRESPONDENCE_NETWORK node table on the physiological side lists only “breathing rate / breathing pattern | wearable | most directly coupled with rhythm,” lacking measurement nodes or edges such as “follow-up sampling design / change-point estimation / encryption window / coarse scan error.” The EDGE_TABLE_science-yixue.md concept extraction table also lacks entries for “inflection point / sampling design / encryption window.” This topic has been registered in science-yixue-l1-gaps.md.


    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.

  • 5 个点里到底有几个独立点:φ 轨迹的 lag-1 自相关有多高,以及它把 MDBC 表整体放宽几成

    Evidence level: C

    一句话核心命题:5 个随访点 ≠ 5 个独立观测——φ 是逐次呼吸算出的统计量,同一场记录内的呼吸本就高度相关,随访间又共享体位、时段、操作员这些慢变量;自相关把有效样本数从 5 打回 2-3,整张 MDBC 表就得按量级放宽,而不是微调。

    正文

    一、问题:整张表都建在一个没人验过的假设上

    丹田呼吸的纵向研究常用 0、1、3、6、12 个月五点随访方案,每次测呼吸相位离散度 φ。上一篇文章(#1184)用这 5 个点算出了「最小可检测拐点位移」(MDBC)表,每一格写着”这个幅度、这个时间点的位移,最少要几次随访才能分辨”。整张表都默认了一件事:5 个随访点 = 5 个独立观测。

    这个假设恰恰最可疑。φ 不是”点一下出一读数”的简单测量——它是从一段连续呼吸记录里算出来的统计量,本身已内含几十上百次呼吸的信息。同一次记录内的呼吸之间本就高度相关(这一口和上一口只差几秒,体位、潮气量几乎没变),不同随访的 φ 又共享同一批慢变量:同一把椅子、同一个时段、同一个操作员。相关性一高,”5 个点”的名义信息量就要大打折——不是微调,是量级上的放宽。

    二、机制:自相关把独立样本数打成几分之几

    “自相关使有效样本数小于名义样本数”有明确的量化框架。一组 n 个观测、lag-1 自相关为 ρ 的序列,有效样本数近似为:

    n_eff = n × (1 − ρ) / (1 + ρ)
    

    把 n=5 代进去:ρ=0 时 n_eff=5(假设成立);ρ=0.3 时 n_eff≈2.7;ρ=0.5 时 n_eff≈1.7;ρ=0.7 时 n_eff≈0.9。只要随访点间 lag-1 自相关到 0.3-0.5,5 个点的有效信息就只有 2-3 个独立观测。

    对生理信号时序做相关性检验的研究反复确认:不考虑自相关会系统性高估信息量、低估置信区间宽度(PMID 39846001)。纵向试验设计里,自相关结构直接改写样本量公式——Kirby 等用 γ^s·θ 双参数相关模型证明:重复测量的自相关越强,达成同样功效所需的等效独立样本越少,忽略自相关会让功效预算系统性乐观(PMID 7913674)。更晚近的整群随机试验方法学把这条推得更细:个体自相关是功效公式里的显式输入项,必须单独建模(PMID 32133688);估计这些相关参数还有专门教程,且多数已发表试验根本不上报这些参数(PMID 37196320)——”5 个点算 5 个独立观测”的普遍做法,几乎没经过自查。

    三、但这不能拿文献系数直接代入

    那去查文献找典型 lag-1 自相关代进去算,行吗?不行。呼吸变异的相关结构并非单一常系数。对感觉运动变异做时间结构分析发现:简单短期结构(相邻观测相关)重测性高、接近稳定特质,而复杂长期记忆成分重测性差(PMID 36618326)。”这 5 个点之间的相关”既含稳定的短程相关、又含飘忽的长程成分,对有效样本量的贡献完全不同。用文献里任何”典型值”代入,都是在用一个未必适用本人群、本协议、本指标的常数,去决定一整张表要不要放宽——这正是本主题必须实测的原因。

    四、标定设计:一次零额外招募的纯相关结构测定

    标定与 #1184 主实验共用受试者,零额外招募:

    1. ≥40 人按 0/1/3/6/12 月五点各测一次 φ(每次 ≥10 分钟);
    2. 逐人拟合 AR(1) 取 lag-1 ρ,算 n_eff = 5 × (1 − ρ) / (1 + ρ);
    3. 把 n_eff 代回 #1184 仿真框架,重算 6-12 月段拐点区间半宽,与按名义 n=5 的结果比对,得放宽倍数。

    小样本下区间估计方法本身会偏(PMID 19010644),故 n_eff 扣减后须用自助法重算区间。

    五、三档判据

    判据一(独立假设崩塌):若 lag-1 中位数 >0.4 且 n_eff 中位数 <3,则”5 点 = 5 独立观测”不成立,整张 MDBC 表须按 n_eff 重算并整体放宽,此前任何按名义点数做的功效预算一律作废。

    判据二(表可用但须附注):若 lag-1 落在 0.1-0.4、n_eff 落在 3-4,则表保留,但每格须附 n_eff 与放宽倍数,报结论时只能报”落在哪个可分辨区间”,不能报具体月数。

    判据三(担忧撤销):若 lag-1 中位数 <0.1 且 n_eff >4,则独立假设近似成立,原表可原样使用,本文担忧全部撤销。

    六、对易学—科学桥接的意义

    丹田呼吸研究要回答”长期修炼到底在生理上改变了什么”,这依赖纵向随访的可靠性,而可靠性又依赖”每个随访点贡献多少独立信息”。若 5 个点实际只有 2-3 个独立观测,此前按 5 点算出的”拐点出现时间””最小可检测位移”都要重新标定——这不是否定丹田呼吸研究,而是把它从”做了 5 次测量”的乐观拉回”只有 2-3 个独立证据”的现实。测量学这一步不补,后面所有”几年练出了什么”的结论都站不稳。

    证据与局限

    已验证: 自相关使时序数据的有效样本数小于名义样本数,且有成熟的 n_eff 近似公式(PMID 39846001);重复测量自相关强度直接改写纵向试验的样本量公式(PMID 7913674);个体自相关是纵向整群试验功效计算的显式输入项(PMID 32133688);估计这些相关参数已有标准教程(PMID 37196320)。

    部分验证: 呼吸/感觉运动变异的时间结构重测性高、但复杂长记忆成分重测性差(PMID 36618326),提示相关结构非单一常系数——这是”必须实测、不能借文献”的依据,但该研究并非丹田呼吸人群的 φ 轨迹数据。

    待验证: 丹田呼吸 φ 轨迹在 0/1/3/6/12 月五点方案下的 lag-1 自相关真实值(本文全部判据都建立在这个尚未测过的数字上);n_eff 扣减后 6-12 月段区间半宽的放宽倍数;小样本下自助法重算的区间是否稳定。

    相关阅读

    证据等级: C(方法学设计裁决:机制有文献支撑,但判据所依赖的 φ 轨迹自相关实测值尚未测出,属测量学标定类探索)
    参考文献: PMID 39846001, PMID 7913674, PMID 32133688, PMID 37196320, PMID 36618326, PMID 19010644, PMID 40119099

    L1 网络状态: 该主题尚未进入 L1 对应网络——「丹田参数 ↔ 测量学维度」簇目前无任何以”随访点 lag-1 自相关 / 有效样本数 n_eff / 独立观测假设”为源或目标的边,已登记入 science-yixue-l1-gaps.md(簇第 19 条)。


    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。

    How many independent points are there among the five: what is the lag-1 autocorrelation of the φ trajectory, and by what factor does it relax the entire MDBC table?

    Evidence level: C

    Core proposition in one sentence: 5 follow-up points ≠ 5 independent observations—φ is a statistic computed from successive breaths; breaths within the same recording are inherently highly correlated, and follow-ups share slow variables such as posture, time of day, and operator. Autocorrelation reduces the effective sample size from 5 back to 2–3, so the entire MDBC table must be scaled up by order of magnitude rather than fine-tuned.

    Main Text

    I. Problem: The Entire Table Rests on an Unverified Assumption

    Longitudinal studies of dantian breathing commonly use a five-point follow-up schedule at 0, 1, 3, 6, and 12 months, measuring the breathing phase dispersion φ each time. The previous article (#1184) used these 5 points to compute the “Minimum Detectable Change in Breathing Phase” (MDBC) table, where each cell states “this magnitude, at this time point, requires at least this many follow-ups to distinguish.” The entire table assumes: 5 follow-up points = 5 independent observations.

    This assumption is precisely the most suspicious part. φ is not a simple measurement that yields one reading per tap—it is a statistic derived from a continuous breathing record, inherently incorporating information from dozens or hundreds of breaths. Breaths within the same recording are already highly correlated (each breath differs from the previous by only a few seconds, with posture and tidal volume nearly unchanged). Moreover, φ values across different follow-ups share the same batch of slow variables: the same chair, the same time slot, the same operator. High correlation means the nominal information content of “5 points” is heavily discounted—not fine-tuned, but scaled down by orders of magnitude.

    II. Mechanism: Autocorrelation Reduces the Effective Sample Size to a Fraction

    “The effective sample size is less than the nominal sample size due to autocorrelation” has a clear quantitative framework. For a sequence of n observations with lag-1 autocorrelation ρ, the effective sample size is approximately:

    n_eff = n × (1 − ρ) / (1 + ρ)
    

    Substituting n = 5: if ρ = 0, n_eff = 5 (the assumption holds); if ρ = 0.3, n_eff ≈ 2.7; if ρ = 0.5, n_eff ≈ 1.7; if ρ = 0.7, n_eff ≈ 0.9. Once the lag-1 autocorrelation between follow-up points reaches 0.3–0.5, the effective information from 5 points amounts to only 2–3 independent observations.

    Studies on temporal correlation in physiological signals repeatedly confirm: ignoring autocorrelation systematically overestimates information content and underestimates confidence interval width (PMID 39846001). In longitudinal trial design, the autocorrelation structure directly rewrites the sample size formula—Kirby et al. used a two-parameter correlation model γ^s·θ and proved: the stronger the autocorrelation of repeated measures, the fewer equivalent independent samples are needed to achieve the same power; ignoring autocorrelation makes the power budget systematically optimistic (PMID 7913674). More recent cluster randomized trial methodology refines this further: individual autocorrelation is an explicit input in the power formula and must be modeled separately (PMID 32133688); there are dedicated tutorials for estimating these correlation parameters, yet most published trials do not report these parameters (PMID 37196320)—the common practice of “calculating 5 independent observations from 5 points” has almost never been self-checked.

    III. But We Cannot Directly Plug Literature Coefficients

    So, why not look up typical lag-1 autocorrelation values in the literature and plug them in? No. The correlation structure of breathing variability is not a single constant coefficient. Time-structure analysis of somatosensory-motor variability reveals: simple short-term structure (adjacent observations) has high test-retest reliability and approximates stable traits, whereas complex long-memory components have poor test-retest reliability (PMID 36618326). “The correlation among these 5 points” includes both stable short-range correlations and fluctuating long-range components, contributing differently to the effective sample size. Using any “typical value” from the literature to decide whether the entire table must be scaled is like applying a constant that may not apply to this specific population, protocol, or metric—this is precisely why this topic requires empirical calibration.

    IV. Calibration Design: Pure Autocorrelation Structure Measurement with Zero Additional Recruitment

    The calibration shares subjects with the main experiment in #1184, requiring zero additional recruitment:

    1. ≥40 participants each measure φ at the five points (0/1/3/6/12 months), with each measurement lasting ≥10 minutes;
    2. Fit an AR(1) model for each individual to obtain lag-1 ρ, then compute n_eff = 5 × (1 − ρ) / (1 + ρ);
    3. Plug n_eff back into the #1184 simulation framework to recompute the half-width of the inflection interval for the 6–12 month segment, comparing it with the nominal n = 5 result to obtain the broadening factor.

    Under small sample sizes, interval estimation methods themselves are biased (PMID 19010644), so after subtracting n_eff, bootstrap resampling must be used to recompute intervals.

    V. Three-tier Criteria

    Criterion 1 (Collapse of the independence assumption): If the median lag-1 autocorrelation > 0.4 and the median n_eff < 3, then “5 points = 5 independent observations” fails; the entire MDBC table must be recalculated using n_eff and scaled up accordingly, invalidating any power budgets previously made based on nominal point counts.

    Criterion 2 (Table usable but with annotations): If lag-1 falls between 0.1–0.4 and n_eff lies between 3–4, the table remains valid but each cell must include n_eff and the broadening factor; conclusions reported can only state “falls within which detectable interval,” not specific months.

    Criterion 3 (Concerns withdrawn): If the median lag-1 autocorrelation < 0.1 and n_eff > 4, then the independence assumption approximately holds; the original table can be used as-is, and all concerns are withdrawn.

    VI. Significance for Bridging Yijing Studies and Science

    Dantian breathing research aims to answer “what physiological changes result from long-term cultivation,” which depends on the reliability of longitudinal follow-ups, in turn depending on “how much independent information each follow-up point contributes.” If the 5 points actually yield only 2–3 independent observations, previously computed “inflection time” and “minimum detectable displacement” must be recalibrated—this does not negate dantian breathing research but pulls it back from the optimism of “5 measurements” to the reality of “only 2–3 independent pieces of evidence.” Without this metrological step, all subsequent conclusions about “what was achieved after several years” will be unstable.

    Evidence and Limitations

    Verified: Autocorrelation reduces the effective sample size of temporal data below the nominal sample size, with established n_eff approximation formulas (PMID 39846001); the strength of autocorrelation in repeated measures directly rewrites the sample size formula for longitudinal trials (PMID 7913674); individual autocorrelation is an explicit input in power calculations for longitudinal cluster trials (PMID 32133688); standard tutorials exist for estimating these correlation parameters (PMID 37196320).

    Partially verified: The test-retest reliability of breathing/somatosensory-motor variability time structures is high, but complex long-memory components have low reliability (PMID 36618326), indicating the correlation structure is not a single constant coefficient—this is the basis for “must be measured empirically, cannot borrow from literature,” though this study does not provide φ trajectory data for the dantian breathing population.

    To be verified: The true lag-1 autocorrelation values of the dantian breathing φ trajectory under the 0/1/3/6/12 month five-point schedule (all criteria in this paper depend on this yet-to-be-measured number); the broadening factor of the interval half-width for the 6–12 month segment after n_eff subtraction; whether bootstrap-resampled intervals remain stable under small samples.

    Related Reading

    Evidence level: C (methodological design decision: the mechanism is supported by literature, but the φ trajectory autocorrelation values required for the criteria have not yet been measured; this belongs to metrological calibration exploration).
    References: PMID 39846001, PMID 7913674, PMID 32133688, PMID 37196320, PMID 36618326, PMID 19010644, PMID 40119099

    L1 network status: This topic has not yet entered the L1 corresponding network—the “dantian parameters ↔ metrological dimensions” cluster currently has no edges sourced from or targeting “follow-up point lag-1 autocorrelation / effective sample size n_eff / independent observation assumption,” and has been registered in science-yixue-l1-gaps.md (cluster item 19).


    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.

  • 「修好了但其实记错了账」到底占几成:残差被吸收后,参数值随瞬时血压漂移的发生率

    Evidence level: C

    残差压到噪声底 ≠ 模型修好了:如果参数值跟着瞬时血压一起漂,那不是修好了,是把结构缺失的方差记到了假参数名下。

    正文

    中医脉象的现代建模里,有一个非常诱人的里程碑:波形残差终于被压到了噪声底。残差小了,模型”拟合好了”,于是顺理成章地,从模型里反推出来的血管参数(硬度、阻力、几何)就被当作”这个人的真实状态”拿来个体化、拿来诊断、拿来随访。这套逻辑在集总参数模型的框架里尤其流行——把动脉树压缩成几个电阻-电容元件,波形一拟合,参数就出来了。

    但 #1179 那一刀切在了最痛的地方:残差压平了,可能只是把账做平了,账本身却记错了。当模型少了一个真实的物理维度(比如血压依赖的硬度变化),拟合算法会把缺掉的那部分方差”吸收”进现有的参数里——参数值因此不再是它本来的生理量,而是”残差吸收器”:一个把结构缺失伪装成个体差异的假变量。

    这一格目前完全靠间接推断。我们知道幅值依赖这个维度真实存在:血压波形的谐波失真(harmonic distortion, HD)随收缩压线性下降(PMID 35433650),交感神经可以独立于血流动力学调节大动脉硬度(PMID 33035441)——这两条证据都指向同一个事实:血管系统里有一个”随血压变化的旋钮”。但没有任何一条证据告诉我们:这个旋钮转动的时候,模型里那 8-10 个参数会跟着漂多少、漂几个、占多大比例。

    而这件事的分量,恰恰藏在”占几成”这三个字里。如果第三格的发生率很高——比如超过三成参数呈现跨个体同号的血压斜率——那么既往所有”单参数个体化成功”的案例都要重判:它们可能集体落在第三格里,报出来的”个体化参数”其实是”当时那一下血压的快照”。

    为什么参数会漂?三条生理路径,层层递进。第一条:硬度本身就是血压的函数。动脉壁的应力-应变关系是非线性的,血压一高,管壁被撑紧,硬度上升,模型里的弹性参数必须跟着变才能继续拟合波形——这在小鼠主动脉上被直接测过:谐波失真与收缩压的线性回归斜率在年龄和饮食组间显著不同(PMID 35433650)。第二条:阻力不是常数。小动脉里的血液是剪切变稀的非牛顿流体,灌注压下降时局部血管阻力反而上升——阻力这个”旋钮”本身就随灌注压变(PMID 26011432)。第三条:交感张力是独立的开关。肌肉交感神经活动(MSNA)可以独立于血压调节大动脉硬度,这意味着即使血压没变,参数也可能在漂(PMID 33035441)。

    更麻烦的是第四条,方法学层面的:当参数之间存在相关结构时,”多解”会让漂移更难识别。贝叶斯校准的全局敏感性分析已经证明,在相关参数结构下,同一组波形可以对应多组同样”拟合良好”的参数组合——你看到的”这个参数变了”,可能只是算法在相关结构里换了一条等效路径(PMID 41838759)。三条生理路径负责”为什么漂”,这一条负责”为什么我们看不出它在漂”。

    那么怎么测?设计其实不复杂,可以和 #1179 主实验共用受试者与波形:≥30 人,每人采集 ≥20 分钟连续波形并同步逐拍血压(指套或连续无创),按 30 秒窗滑动重估参数,得到每个参数的”时间序列”。然后做三件事:① 对每个参数做”参数值 vs 瞬时收缩压”的混合效应回归,斜率显著且跨个体同号者判为”系统性漂移”;② 只在残差已被压到噪声底的那些阶次上统计漂移发生率——这一条最重要,它保证我们测的不是”拟合不好所以参数乱跳”,而是”拟合好了参数还在跟着血压跑”;③ 用隔日重测的稳定性作对照,区分”真稳定的个体化值”与”看起来稳定其实在漂”的假稳定。

    判据分三档。若在残差已达噪声底的阶次中,有超过 30% 的参数呈现跨个体同号的显著血压斜率——即漂移是系统性的而非噪声——那么”残差压到噪声底”就不足以支撑”小修成功”的结论,第二格判据必须附加”参数稳定性”条件,既往所有单参数个体化结论整体降级。若在 10%-30% 之间,则须在报告个体化参数时强制附带”该参数在本受试者上的血压斜率及其置信区间”。若低于 10% 且斜率符号随机,则第三格降为次要情形,小修判据维持原样。

    这件事对易学-科学桥接的意义在于:脉象是”窗口”,但窗口的玻璃会随气压变形。当我们说”这个人的脉象参数反映了某条经络/某个脏腑的状态”时,如果参数本身是血压快照而非稳定个体特征,那么跨时辰、跨日、跨状态的比较就全都不成立——时辰对应关系(十二时辰与五脏)恰恰建立在”同一个人不同时间的脉象差异”之上。参数漂移率实测,就是给这张”时辰-脏腑对应网”做地基勘察:如果地基在随血压晃动,上面的对应关系就要重新标定。

    L1 网络状态:CORRESPONDENCE_NETWORK 边表止于 E-050(纳音↔频率,open),无任何以”参数值随瞬时血压漂移 / 残差吸收 / 血压依赖硬度 / 假参数”为源或目标的边;EDGE_TABLE 概念计数中”谐波 147 篇/A 级 20″仅覆盖波形谐波失真与硬度(PMID 35433650 等),无”参数漂移发生率”概念条目。该主题尚未进入 L1 对应网络,已登记 gap(谐波残差簇第 19 条)。

    证据与局限

    已验证: 血压波形谐波失真随收缩压线性变化,幅值依赖维度真实存在(PMID 35433650);交感神经可独立于血流动力学调节大动脉硬度(PMID 33035441);微血管阻力随灌注压变化、非恒定(PMID 26011432)。

    部分验证: 相关参数结构下多解使”参数漂移”更难识别(PMID 41838759,方法学仿真层面)。

    待验证: ① ≥30 人连续波形+逐拍血压下,残差压到噪声底的阶次中参数随血压漂移的实际发生率(第三格占比);② 隔日重测能否区分”真稳定个体化值”与”假稳定”;③ 漂移率>30% 时,既往单参数个体化结论的实际降级比例。

    相关阅读

    证据等级: C(传统文化共识/初步探索——核心问题”漂移发生率”无直接实证,间接证据为动物实验与方法学仿真)
    参考文献: PMID 35433650, PMID 33035441, PMID 26011432, PMID 41838759


    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。

    How much of \”Fixed but Recorded Wrong\” is accounted for: the incidence rate of parameter drift following residual absorption in instantaneous blood pressure

    Evidence level: C

    Residuals compressed to the noise floor ≠ model fixed: if parameter values drift with instantaneous blood pressure, it is not fixed; the missing structural variance has been recorded under false parameter names.

    Body

    In modern modeling of Traditional Chinese Medicine pulse waveforms, there is a particularly alluring milestone: waveform residuals have finally been compressed to the noise floor. Residuals are small, so the model “fits well”; consequently, vascular parameters (stiffness, resistance, geometry) inferred from the model are treated as “the individual’s true state” for personalization, diagnosis, and follow-up. This logic is especially prevalent within the lumped-parameter model framework—where the arterial tree is compressed into a few resistor-capacitor elements, and once the waveform is fitted, parameters emerge.

    But #1179 cuts at the most painful point: residuals flattened may only balance the account, not record it correctly. When the model lacks a true physical dimension (e.g., stiffness variation dependent on blood pressure), the fitting algorithm absorbs the missing variance into existing parameters—parameter values are no longer their original physiological quantities but become “residual absorbers”: false variables that disguise structural deficits as individual differences.

    This aspect currently relies entirely on indirect inference. We know amplitude dependence is a real dimension: harmonic distortion (HD) of the blood pressure waveform decreases linearly with systolic pressure (PMID 35433650), and the sympathetic nervous system can independently regulate large artery stiffness via hemodynamic regulation (PMID 33035441)—both pieces of evidence point to the same fact: there is a “blood-pressure-dependent knob” in the vascular system. Yet no evidence tells us how much, how many, or what proportion of the model’s 8–10 parameters drift when this knob turns.

    The weight of this matter lies precisely in the phrase “what proportion.” If the incidence rate of the third scenario is high—for example, if more than 30% of parameters exhibit cross-individual consistent blood pressure slopes—then all previous cases of “successful single-parameter personalization” must be re-evaluated: they may collectively fall into the third scenario, where reported “personalized parameters” are actually “snapshots of that moment’s blood pressure.”

    Why do parameters drift? Three physiological pathways, progressing stepwise. First: stiffness itself is a function of blood pressure. The stress-strain relationship of the arterial wall is nonlinear; as blood pressure rises, the vessel wall stretches tighter, increasing stiffness, so the elastic parameter in the model must adjust to continue fitting the waveform—this has been directly measured on mouse aortas: the linear regression slope of harmonic distortion versus systolic pressure differs significantly between age and diet groups (PMID 35433650). Second: resistance is not constant. Blood in arterioles behaves as a shear-thinning non-Newtonian fluid; when perfusion pressure drops, local vascular resistance actually increases—the “resistance knob” itself varies with perfusion pressure (PMID 26011432). Third: sympathetic tone is an independent switch. Muscle sympathetic nerve activity (MSNA) can regulate large artery stiffness independently of blood pressure, meaning even if blood pressure remains unchanged, parameters may still drift (PMID 33035441).

    More troubling is the fourth aspect, methodological: when parameters exhibit correlated structure, “multiple solutions” make drift harder to identify. Bayesian calibration global sensitivity analysis has shown that under correlated parameter structures, the same waveform can correspond to multiple equally well-fitting parameter combinations—what you see as “this parameter changed” may simply be the algorithm switching to an equivalent path within the correlated structure (PMID 41838759). Three physiological pathways explain “why it drifts”; this one explains “why we cannot see it drifting.”

    How to measure? The design is not complex and can share subjects and waveforms with the main #1179 experiment: ≥30 individuals, each collecting ≥20 minutes of continuous waveform with synchronized beat-by-beat blood pressure (finger cuff or continuous non-invasive), sliding re-estimation of parameters in 30-second windows to obtain “time series” for each parameter. Then three steps: ① mixed-effects regression of “parameter value vs instantaneous systolic pressure” for each parameter; significant slopes with cross-individual consistent signs indicate “systematic drift”; ② calculate drift incidence rate only for orders where residuals have been compressed to the noise floor—this is crucial, ensuring we measure “parameters still drifting despite good fit” rather than “poor fit causing erratic parameters”; ③ use cross-day retest stability as control to distinguish “truly stable personalized values” from “apparently stable but actually drifting” false stability.

    Criteria are divided into three tiers. If, among orders where residuals have reached the noise floor, more than 30% of parameters show cross-individual consistent significant blood pressure slopes—i.e., drift is systematic rather than noise—then “residuals compressed to the noise floor” is insufficient to support the conclusion of “minor repair success”; the second-tier criterion must add a “parameter stability” condition, and all previous single-parameter personalization conclusions must be downgraded collectively. If between 10%–30%, then reporting personalized parameters must mandatorily include “the blood pressure slope and confidence interval for this parameter in this subject.” If below 10% with random slope signs, the third scenario becomes secondary, and the minor repair criterion remains unchanged.

    The significance of this for bridging Yijing Studies and science lies in: pulse waveforms are “windows,” but the glass of the window deforms with air pressure. When we say “this person’s pulse parameter reflects the state of a meridian or organ,” if the parameter itself is a blood pressure snapshot rather than a stable individual characteristic, then comparisons across hours, days, and states all become invalid—the hour correspondence (Twelve Hours and Five Organs) is precisely built upon “pulse differences of the same person at different times.” Measuring the parameter drift rate is essentially surveying the foundation of this “Hour-Organ correspondence network”: if the foundation sways with blood pressure, the corresponding relationships above must be recalibrated.

    L1 Network Status: CORRESPONDENCE_NETWORK edge table ends at E-050 (Na Yin ↔ Frequency, open); no edges originate from or target “parameter drift following instantaneous blood pressure / residual absorption / blood pressure-dependent stiffness / false parameters”; EDGE_TABLE concept count includes “harmonic 147 papers / Level A 20” covering only waveform harmonic distortion and stiffness (PMID 35433650, etc.), but lacks the “parameter drift incidence rate” concept entry. This topic has not yet entered the L1 correspondence network; gap registered (Harmonic Residual Cluster #19).

    Evidence and Limitations

    Verified: Blood pressure waveform harmonic distortion varies linearly with systolic pressure, confirming the amplitude-dependent dimension exists (PMID 35433650); sympathetic nervous system can independently regulate large artery stiffness via hemodynamic regulation (PMID 33035441); microvascular resistance varies with perfusion pressure and is not constant (PMID 26011432).

    Partially Verified: Multiple solutions under correlated parameter structures make “parameter drift” harder to identify (PMID 41838759, methodological simulation level).

    Pending Verification: ① actual incidence rate of parameter drift following blood pressure among orders where residuals are compressed to the noise floor under ≥30 individuals with continuous waveform and beat-by-beat blood pressure (proportion of the third scenario); ② whether cross-day retest can distinguish “truly stable personalized values” from “false stability”; ③ actual downgrade proportion of previous single-parameter personalization conclusions when drift rate exceeds 30%.

    Further Reading

    Evidence Level: C (traditional consensus / preliminary exploration—core question “drift incidence rate” lacks direct empirical evidence; indirect evidence comes from animal experiments and methodological simulation)
    References: PMID 35433650, PMID 33035441, PMID 26011432, PMID 41838759


    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.

  • 「敏感但不准」这一格该怎么记:低灵敏度床的命中要作废还是降权,得看它在判决里占了多少票

    Evidence level: C

    核心命题:低灵敏度床的”命中”该作废还是降权,不是原则问题而是计数问题——它在判决里占的票数,决定处置方式会不会推动结论方向。

    正文

    母文(同源五床体检系列)把”参数扫描谱变化 <5%”从”不可辨识”改判为”可辨识但对输出没影响”之后,留下了一个统计上的半成品:这种”敏感但不准”的床,它的命中到底该怎么记账?原文只写了”保留比对、命中不计为支持证据”,但”不计”与”降权”在数学上是两回事,而这个选择本身会推动结论方向。

    先看机制。一个床的灵敏度,指的是它的出口阻力扫过生理范围时,桡动脉脉搏前几阶谐波变化多大。灵敏度高,观测对它有信息量;灵敏度低,它怎么变,你看到的波形都几乎不动。但”输出看不见它”与”它不存在”是两码事:1-D 血流动力学模型里,血管区域阻力是可以通过拟合辨识出来的(PMID 22771032),只是不同区域的可辨识程度天然不同。全局灵敏度分析也反复显示,模型对某些参数高度敏感、对另一些参数几乎不敏感,而识别出”哪些参数可以固定在名义值”正是灵敏度分析的核心用途之一(PMID 26367184);在肺血流动力学 1-D 模型里,研究者正是先做灵敏度排序、再做相关性分析,才筛出”可辨识参数子集”(PMID 31355521)。

    这里的关键是:低灵敏度不等于没辨识出来,而是”辨识得再准,输出也看不见”。后果不是”无法比对”,而是”它的命中没有信息量”。但问题接踵而至——把这样的床从五床判决里”剔除”,等价于把它的权重设成 0;”降权”则等价于设成一个小正数。两者给出的估计量不同,差的正是这个床在判决里的实际票数。若这个低灵敏度床在既往数据里常常”命中”,剔掉它会让整体命中率大幅下降;若它常常”偏离”,剔掉反而让结论变好看。也就是说,处置方式本身成了推动结论方向的手,而母文没给出”该走哪条”的判据。

    这个问题有现成的统计骨架。参数辨识里有个著名的坑:参数越多越容易过参数化,即多组参数组合能同样好地拟合同一批数据,无法判定哪组才是真实的力学属性(PMID 33934232)——这正是”低灵敏度床”在结构层面的根源,它挤占了本该属于灵敏床的自由度。而在数值侧,Windkessel 模型辨识早就发现:某些参数(如惯性项 L)的估计噪声几乎全部来自它自身、是噪声主源(PMID 20172779),这类参数恰恰是”敏感但不准”的典型。这些证据指向同一个方法论要求:处置低灵敏度床之前,必须先量化它在判决里的票数占比。

    可行的落点是判决敏感性回算(不需要新受试者):用 1-D 模型合成五床各自的灵敏度谱,构造若干组”真值映射 + 噪声”的仿真判决,让每床分别以高/中/低灵敏度出现;对同一批判决分别按三种口径统计命中率——全计数、剔除低灵敏度床、按灵敏度加权。三种口径的命中率极差,就是处置选择的”推动力”上限。这里还有一层方法学细节值得注意:边界条件的设置本身会以不小量级改变仿真结论(PMID 26803339),所以灵敏度谱必须在固定的边界条件假设下合成,否则”低灵敏度”本身就是个随设定漂移的标签——也就是说,在实测灵敏度之前,”低灵敏度床”这一判定本身就可能不可靠,这更说明处置规则不能拍脑袋定,而要按票数占比动态定。

    本文作为方法学提案(纯回算、无新受试者),把”该作废还是降权”从一个措辞问题转译成一个可数的量(三口径命中率极差 + 低灵敏度床票数占比),并给出三条可执行判据:极差 <5 个百分点,处置选择无关紧要,按最省事的”剔除”办;极差 5–15 个百分点,方法里必须预先声明口径并做敏感性分析(两口径都报);极差 >15 个百分点且低灵敏度床票数占比 >20%,”命中率”这个指标本身在五床设计里不可用,须改为逐床分报、不做跨床合计。这样,”敏感但不准”这一格就从”软处理”变成”有数可依的裁决”。

    证据与局限

    已验证: 1-D 血流动力学模型可辨识血管区域阻力(PMID 22771032);全局灵敏度分析可识别并固定不敏感参数(PMID 26367184);过参数化会导致多组参数同解、不可辨识(PMID 33934232)。
    部分验证: 灵敏度排序 + 相关性分析可筛出可辨识参数子集(PMID 31355521);Windkessel 某些参数(L)是噪声主源、敏感但不准(PMID 20172779);边界条件设定会以可观量级改变仿真输出(PMID 26803339)——这些均来自动脉建模文献,未在”五床出口阻力 + 中医脉象谐波”语境下直接实测。
    待验证: 三口径命中率极差在真实五床数据上的实际分布;低灵敏度床的票数占比;边界条件假设对”低灵敏度”判定的漂移幅度。

    相关阅读

    证据等级: C
    参考文献: PMID 22771032, PMID 26367184, PMID 31355521, PMID 20172779, PMID 33934232, PMID 26803339

    L1 网络状态:CORRESPONDENCE_NETWORK 边表止于 E-050(六十甲子纳音↔频率,状态 open,现代证据栏原文”尚无任何可测来源”),无任何以血管床阻力、参数反解、灵敏度谱、加权命中率为源或目标的边——该主题(低灵敏度床命中处置)尚未进入 L1 对应网络(母文 #1246 已登记为该簇第 16 条,本行为其深化层追加登记)。


    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。

    How to remember \”sensitive but inaccurate\”: whether low-sensitivity bed hits are discarded or downweighted depends on how many votes it casts in the verdict.

    Evidence level: C

    Core Proposition: Whether the “hit” of a low-sensitivity bed should be voided or downweighted is not a principled issue but a counting issue—its share of votes in the verdict determines whether its treatment will push the conclusion direction.

    Body

    The parent text (from the homologous five-bed physical examination series) reclassified “parameter scan spectrum variation <5%” from “unidentifiable” to “identifiable but with no impact on output,” leaving behind a statistical semi-finished product: how should the hit of such a “sensitive but inaccurate” bed be recorded? The original text only stated “retain comparison, do not count hits as supporting evidence,” but “not counting” and “downweighting” are mathematically distinct, and this choice itself will push the conclusion direction.

    First, examine the mechanism. A bed’s sensitivity refers to how much the first few harmonic changes of the radial artery pulse wave occur when its outlet resistance sweeps across the physiological range. High sensitivity means observation has information about it; low sensitivity means how it changes, the waveform you see barely moves. But “output cannot see it” is different from “it does not exist”: in the 1-D hemodynamic model, vascular region resistance can be identified through fitting (PMID 22771032), though the identifiability of different regions is naturally unequal. Global sensitivity analysis also repeatedly shows that models are highly sensitive to some parameters and almost insensitive to others, and identifying “which parameters can be fixed at nominal values” is precisely one of the core uses of sensitivity analysis (PMID 26367184); in the lung hemodynamic 1-D model, researchers first performed sensitivity ranking and then correlation analysis to screen out the “identifiable parameter subset” (PMID 31355521).

    The key point here is: low sensitivity does not mean it was not identified, but rather “even if identified perfectly, the output cannot see it.” The consequence is not “cannot compare,” but “its hit carries no information.” Yet problems follow immediately—removing such a bed from the five-bed verdict is equivalent to setting its weight to 0; “downweighting” is equivalent to setting it to a small positive number. The two approaches yield different estimators, differing precisely by the actual number of votes this bed casts in the verdict. If this low-sensitivity bed often “hits” in past data, removing it will drastically reduce the overall hit rate; if it often “misses,” removing it instead makes the conclusion look better. In other words, the treatment method itself becomes the hand that pushes the conclusion direction, yet the parent text provides no criterion for “which path to take.”

    This problem has an existing statistical skeleton. In parameter identification, there is a famous pitfall: the more parameters, the easier over-parameterization occurs, meaning multiple sets of parameter combinations can fit the same batch of data equally well, making it impossible to determine which set represents the true mechanical properties (PMID 33934232)—this is precisely the structural root of “low-sensitivity beds,” as they occupy degrees of freedom that should belong to sensitive beds. On the numerical side, Windkessel model identification long ago discovered: certain parameters (such as inertial term L) have estimation noise almost entirely originating from themselves, making them the primary noise source (PMID 20172779); such parameters are exactly the typical case of “sensitive but inaccurate.” These evidences point to the same methodological requirement: before treating low-sensitivity beds, one must first quantify their vote share in the verdict.

    The feasible landing point is verdict sensitivity back-calculation (no new subjects required): use the 1-D model to synthesize each bed’s sensitivity spectrum, construct several groups of “true value mapping + noise” simulated verdicts, letting each bed appear with high/medium/low sensitivity respectively; then statistically compute hit rates under three perspectives—full counting, excluding low-sensitivity beds, and weighting by sensitivity. The extreme difference in hit rates across these three perspectives represents the upper limit of the “driving force” of treatment choice. Here lies another methodological detail worth noting: boundary condition settings themselves can alter simulated conclusions by a significant magnitude (PMID 26803339), so the sensitivity spectrum must be synthesized under fixed boundary condition assumptions; otherwise “low sensitivity” itself becomes a label drifting with the setting—meaning, before actual measurement of sensitivity, the judgment of “low-sensitivity bed” may already be unreliable, which further proves treatment rules cannot be arbitrarily decided but must be dynamically determined by vote share.

    This paper, as a methodological proposal (pure back-calculation, no new subjects), translates the question of “void or downweight” from a wording issue into a countable quantity (extreme difference in hit rates + low-sensitivity bed vote share), and provides three executable criteria: if the extreme difference is <5 percentage points, treatment choice is irrelevant, proceed with the most convenient “exclusion”; if the extreme difference is 5–15 percentage points, the method must predeclare the perspective and conduct sensitivity analysis (report both perspectives); if the extreme difference exceeds 15 percentage points AND the low-sensitivity bed vote share exceeds 20%, the metric of “hit rate” itself is unusable in the five-bed design, must instead report each bed separately without cross-bed summation. Thus, “sensitive but inaccurate” transforms from a “soft handling” into a “decided-by-numbers ruling.”

    Evidence and Limitations

    Verified: 1-D hemodynamic models can identify vascular region resistance (PMID 22771032); global sensitivity analysis can identify and fix insensitive parameters (PMID 26367184); over-parameterization leads to multiple equivalent solutions and unidentifiability (PMID 33934232).

    Partially Verified: Sensitivity ranking + correlation analysis can screen out identifiable parameter subsets (PMID 31355521); certain Windkessel parameters (L) are primary noise sources, sensitive but inaccurate (PMID 20172779); boundary condition settings can alter simulated outputs by significant magnitude (PMID 26803339)—all derived from arterial modeling literature, not directly measured in the “five-bed outlet resistance + traditional pulse wave harmonics” context.

    To Be Verified: Actual distribution of extreme hit rate differences on real five-bed data; vote share of low-sensitivity beds; drift amplitude of “low sensitivity” judgment due to boundary condition assumptions.

    Further Reading

    Evidence Level: C
    References: PMID 22771032, PMID 26367184, PMID 31355521, PMID 20172779, PMID 33934232, PMID 26803339

    L1 Network Status: CORRESPONDENCE_NETWORK edge list stops at E-050 (Sixty Jiazi Yin-Ying ↔ Frequency, status open, modern evidence column original text “no measurable source yet”), with no edges sourced from or targeting vascular bed resistance, parameter inversion, sensitivity spectrum, weighted hit rate—this topic (low-sensitivity bed hit treatment) has not yet entered the L1 corresponding network (parent text #1246 registered this cluster’s 16th item, this line is its deepening layer addition registration).


    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.

  • 五个床到底有几个自由度:并联出口拓扑的代数可辨识性计数,与「第五个床」是不是从来就没被定住过

    Evidence level: C

    桡动脉只给总流量与总压力,五个器官出口阻力若以并联方式合成,那么总阻力被观测定住之后,五个床里到底还剩几个可以独立变动的参数——这个数,必须先算清,再谈逐器官体检。

    正文

    中医脉诊把桡动脉的搏动当作「五脏六腑的窗口」:寸关尺三部各配脏腑,一根血管上读出五六个器官的信息。现代的脉象研究把它转译成一套工程问题——把五个器官的出口阻力(outflow resistance)当作五个并联的「床」(bed),通过主动脉与桡动脉的压力-流量波形,反解出每个床的阻力值,从而判断哪个器官的血流阻力偏高。这个思路在物理上说得通:心输出量分配到各器官,阻力大的器官拿到的份额小,波形的谐波结构里就带着份额的信息(PMID 21622820 的 1D 模型分类实验表明,波形谐波确实携带可被机器识别的生理差异)。

    问题出在反解的数学结构上。五个床以并联方式汇入一根桡动脉,桡动脉这个测量点只能给出一个总流量和一个总压力。如果五个床的阻力以并联方式合成总阻力,那么总阻力一旦被观测定住,五个阻力值之间就只剩四个可独立变动的比值——第五个由其余四个自动决定。换句话说,「五个床全部可辨识」这一格,在设计上可能本来就是空集。这不是中医理论的问题,是所有并联汇流结构的通病:测量点在上游、分支点在下游,下游的每个分支对上游信号的贡献被混叠在一起,能分开多少,取决于模型里给每个分支装了几个参数。

    这一步不能凭直觉下结论。并联合成到底吞掉几个自由度,取决于模型里还带不带各支路的惯性(L)与顺应性(C)。如果每个床只装一个纯电阻 R,五个床就是五个参数对两条观测约束,自由度为五减二等于三——只有三个能定住。如果每个床带 R 与 C 两个参数,五个床就是十个参数对两条观测约束,即使每条支路的波形形状信息都算进去,自由度是否够到十也必须算清,不能凭「有相位关系」一句带过。

    这个问题在心血管建模领域有现成的判据。Kirk 等人用微分代数工具 DAISY 对几类心血管模型做了先验全局可辨识性分析:三元件 Windkessel、四元件 Windkessel、T-Tube 模型逐一判定,结果是所有模型都全局可辨识,唯独 T-Tube 这个「一入口两出口」的拓扑被判为非全局可辨识(PMID 26726299)。T-Tube 与我们的问题同构:入口在上游、两个出口在下游,下游出口的参数不能从上游测量唯一反解。五个床比两个出口更极端——出口越多,可辨识性越差。另一条单管证据来自参数估计的实践:四元件 Windkessel 模型里,惯性参数 L 的估计误差系统性偏大,随着噪声水平升高,估计结果的离群值几乎全部来自 L 的误差(PMID 20172779)。这提醒我们,即使结构上可辨识的参数,在数值上也分三六九等——有的参数天生「软」,一有噪声就飘。

    可辨识性不是纯数学游戏,它直接决定实验设计能不能做。Blanco 等人在一个解剖精细的手臂模型中——67 段动脉、16 个血管床——用 Broyden 法反解各床阻力,成功的前提是他们对「哪几个参数可辨识」做了敏感性分析(PMID 22771032)。反过来说,如果五个床的结构可辨识维数不足,那么再多的被试、再多的波形采样也救不回来:测量学上有个铁律,样本量解决的是随机误差,结构不可辨识是系统性错误,加样本只会把错误的估计「精确地」定在错的值上。

    要裁决这件事,需要三步。第一步是纯代数判定:用微分代数方法(DAISY 一类工具,PMID 26726299 的现成路径)对五个具体拓扑逐个跑先验全局可辨识性——五床纯 R、五床 R+C、五床 R+C+L 三种配置各判一次,得到「结构上可辨识的参数个数」确切值。第二步是数值侧验证:用 1D 模型(PMID 22771032 的 67 段 16 床手臂模型是可复用的骨架)做多初值反解,统计实际收敛到唯一解的维数是否等于代数判定值。第三步是关键的一步——降维检验:把五床逐步合并成四床、三床,看谱误差怎么变。若合并到某床数时误差才陡然上升,说明前面那些床本来就是靠共线凑出来的,多出来的「床」并没有携带独立信息。

    三种可能的结果各有处置。若代数判定显示五床 R+C 配置下可辨识维数小于十(即至少一床结构上不可辨识),那么逐器官裁决必须改写为「可辨识子空间内的裁决」,「五床全可辨识」这一格要删去,它的样本量不必再凑。若五床纯 R 配置恰好可辨识、而加 C 后不可辨识,那么模型必须明确声明「只用 R」,任何带顺应性的版本都要另立判据。若三种配置下可辨识维数都等于参数个数,那么「并联约束吞自由度」的推论被推翻,五床体检按原计划做。

    对中医脉诊的桥接意义在于:古人说「脉有胃气」、说「五脏脉各有所主」,但没说清楚一根桡动脉到底能装下几条独立的信息流。现代测量学给了我们一个清醒的答案方向——一根血管的信息容量是有限的,五个床能定住几个、哪几个,是可以先算出来再上临床的。这比「五个都能测」的乐观假设更接近传统脉诊的实际:历代医家强调「舍脉从证」「脉证参合」,恰恰是因为脉本身的信息维度不够,需要证来补。

    证据与局限

    已验证: T-Tube(一入口两出口)拓扑在先验全局可辨识性分析中是唯一非全局可辨识的心血管模型(PMID 26726299);四元件 Windkessel 中惯性参数 L 的估计误差系统性偏大、离群值几乎全部来自 L(PMID 20172779);67 段 16 床手臂模型的阻力反解可行、但以敏感性分析为前提(PMID 22771032)。
    部分验证: 「并联合成吞自由度」的推论在纯电阻与 R+C 配置下的确切维数尚属代数推论,未经五床具体拓扑的 DAISY 判定;1D 模型多初值反解的实际收敛维数与代数判定值的对照尚无数据。
    待验证: ① 五床 R / R+C / R+C+L 三种配置的先验全局可辨识性维数(DAISY 判定);② 多初值反解收敛维数与代数判定值的一致性;③ 五床逐步合并四床/三床时的谱误差陡升点位置——该位置决定「真实床数」是几。

    L1 网络状态:对应边 E-001~E-005(五行↔五脏)在 CORRESPONDENCE_NETWORK.md 中全部标注 open(待验证),现代证据列「待补」,L2 以上边数 = 0;边表显示「五脏」概念最高证据级 B(A 级 0 篇)、「脉象」的 A 级证据来自波形信号处理本身而非「脉象↔五脏」对应边。本文涉边为「脉象波形 → 器官阻力反解」,在 L1 中属「薄 / 仅文献」状态,故证据等级标 C。

    相关阅读

    证据等级: C
    参考文献: PMID 26726299, PMID 20172779, PMID 22771032


    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。

    How Many Degrees of Freedom Do Five Beds Actually Have: Algebraic Identifiability Counting for Parallel Export Topology, and Was the \”Fifth Bed\” Never Fixed?

    Evidence level: C

    The radial artery provides only total flow and total pressure. If the outflow resistances of five organs are synthesized in parallel, then once the total resistance is observed and fixed, how many independent parameters remain among the five beds? This number must be calculated first before proceeding to organ-by-organ assessment.

    Body

    Traditional Chinese medicine pulse diagnosis treats the pulsation of the radial artery as a “window into the five viscera and six fu-organs”: the Cun-Guan-Chi sections each correspond to specific organs, allowing five or six organ signals to be read from a single vessel. Modern pulse research translates this into an engineering problem—treating the outflow resistance of five organs as five parallel “beds,” and using pressure-flow waveforms from the aorta and radial artery to inversely solve each bed’s resistance value, thereby identifying which organ exhibits elevated blood flow resistance. This approach is physically coherent: cardiac output is distributed among organs, so organs with higher resistance receive a smaller share, and the harmonic structure of the waveform carries information about these shares (the 1D model classification experiment in PMID 21622820 demonstrates that waveform harmonics indeed carry physiologically distinct features recognizable by machines).

    The difficulty lies in the mathematical structure of the inverse solution. Since five beds merge into a single radial artery, this measurement point yields only one total flow and one total pressure. If the five bed resistances combine in parallel to form a total resistance, then once that total resistance is observed and fixed, only four independent ratios remain among the five resistance values—the fifth is automatically determined by the other four. In other words, the proposition that “all five beds are identifiable” may be an empty set by design. This is not a flaw in traditional theory, but a universal limitation of any parallel-converging structure: when the measurement point is upstream and branch points are downstream, contributions from each downstream branch are superimposed in the upstream signal; how much can be separated depends entirely on how many parameters the model assigns to each branch.

    This conclusion cannot rely on intuition alone. How many degrees of freedom are swallowed by parallel synthesis depends on whether the model includes inertia (L) and compliance (C) for each branch. If each bed is modeled as a pure resistor R, then five beds yield five parameters against two observational constraints, leaving five minus two equals three degrees of freedom—only three can be fixed. If each bed includes both R and C, then five beds yield ten parameters against two constraints; even if waveform shape information is fully utilized, whether the degrees of freedom suffice for ten must be calculated precisely, not assumed based on “phase relationships.”

    In cardiovascular modeling, established criteria exist. Kirk et al. used differential-algebraic tools (DAISY) to perform prior global identifiability analysis on several cardiovascular models: three-element Windkessel, four-element Windkessel, and T-Tube models were evaluated sequentially. All were globally identifiable except the T-Tube topology—a “one inlet, two outlets” configuration—which was found to be non-globally identifiable (PMID 26726299). The T-Tube is isomorphic to our problem: the inlet is upstream, two outlets are downstream, and parameters of the downstream outlets cannot be uniquely inverted from upstream measurements. Five beds represent an even more extreme case than two outlets—the more outlets, the worse identifiability. Another single-tube evidence comes from parameter estimation practice: in the four-element Windkessel model, the estimation error of inertia parameter L is systematically larger; as noise level increases, outliers in estimated results almost entirely stem from errors in L (PMID 20172779). This reminds us that even structurally identifiable parameters differ in numerical reliability—some are inherently “soft,” drifting with any noise.

    Identifiability is not merely a mathematical game; it directly determines whether experimental design is feasible. Blanco et al., in an anatomically detailed arm model—67 arterial segments, 16 vascular beds—successfully inverted resistance for each bed using Broyden’s method, but only after performing sensitivity analysis to determine which parameters were identifiable (PMID 22771032). Conversely, if the structural identifiability dimension of five beds is insufficient, no amount of subjects or waveform sampling can rescue the solution: measurement theory dictates that sample size addresses random error, whereas structural non-identifiability is a systematic error; adding samples merely “precisely” locks the wrong estimate.

    To resolve this, three steps are required. First is pure algebraic determination: apply differential-algebraic methods (tools like DAISY, following the established path in PMID 26726299) to evaluate prior global identifiability for each of the five specific topologies—five beds with pure R, five beds with R+C, and five beds with R+C+L—yielding the exact count of structurally identifiable parameters. Second is numerical validation: perform multi-initial-condition inverse solving using 1D models (the 67-segment, 16-bed arm model in PMID 22771032 serves as a reusable framework) and statistically verify whether the actual convergence dimension matches the algebraic prediction. Third—and crucial—is dimensionality reduction testing: progressively merge five beds into four, then three, observing how spectral error evolves. If error sharply rises only when merging to a certain number of beds, it indicates that earlier beds were artificially constructed via collinearity; the extra “beds” carried no independent information.

    Three possible outcomes each dictate a different course of action. If algebraic determination shows that under the R+C configuration the identifiable dimension is less than ten (i.e., at least one bed is structurally non-identifiable), then organ-by-organ adjudication must be rewritten as “adjudication within the identifiable subspace,” and the claim that “all five beds are identifiable” must be deleted; no additional sample size is needed for it. If the pure R configuration happens to be identifiable while adding C renders it non-identifiable, the model must explicitly state “use only R,” and any version incorporating compliance requires its own separate criterion. If the identifiable dimension equals the parameter count across all three configurations, the inference that “parallel constraints swallow degrees of freedom” is overturned, and the five-bed examination proceeds as originally planned.

    The bridging significance for traditional Chinese medicine pulse diagnosis is profound: ancients spoke of “pulse having stomach qi” and “each visceral pulse governs its organ,” yet they did not clarify how many independent information streams a single radial artery can actually hold. Modern measurement science offers a sober direction: the information capacity of one vessel is finite; how many beds can be fixed, and which ones, must be calculated before clinical application. This is closer to the reality of traditional pulse diagnosis than the optimistic assumption that “all five can be measured.” Historically, physicians emphasized “舍脉从证” (relying on symptoms when pulse is inconclusive) and “脉证参合” (combining pulse and symptom), precisely because the pulse itself lacks sufficient informational dimension, requiring symptoms to supplement it.

    Evidence and Limitations

    Verified: The T-Tube (one inlet, two outlets) topology is the only cardiovascular model found to be non-globally identifiable in prior global identifiability analysis (PMID 26726299); in the four-element Windkessel, estimation error of inertia parameter L is systematically larger, with outliers almost entirely attributable to L (PMID 20172779); resistance inversion in the 67-segment, 16-bed arm model is feasible but contingent on sensitivity analysis (PMID 22771032).
    Partially Verified: The exact dimensionality of the “parallel synthesis swallows degrees of freedom” inference remains an algebraic deduction for pure resistor and R+C configurations, pending DAISY evaluation of the five-bed specific topology; comparison between actual convergence dimension from multi-initial-condition inverse solving and algebraic predicted value lacks data.
    To Be Verified: ① Prior global identifiability dimensions for the three configurations of five beds (pure R, R+C, R+C+L) via DAISY; ② Consistency between convergence dimension of multi-initial-condition inverse solving and algebraic prediction; ③ Location of the sharp rise in spectral error during progressive merging of five beds into four/three—this position determines the true number of beds.

    L1 Network Status: Corresponding edges E-001~E-005 (Five Phases ↔ Five Viscera) are all marked open (to be verified) in CORRESPONDENCE_NETWORK.md; modern evidence lists “to be supplemented,” and edges at L2 and above have zero count. Edge table shows the concept of “five viscera” has highest evidence level B (zero Level A papers); Level A evidence for “pulse pattern” stems from waveform signal processing itself, not the correspondence edge between “pulse pattern ↔ five viscera.” The edges involved in this article are “pulse waveform → organ resistance inversion,” which, within L1, fall into the “thin / literature-only” state; hence evidence level is marked C.

    Further Reading

    Evidence Level: C
    References: PMID 26726299, PMID 20172779, PMID 22771032


    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.

  • 认人不认人,是分层最好的试金石:身份层与状态层在跨年识别任务里的表现对比

    Evidence level: B

    一句话核心命题:如果频率签名里真的存在”身份层”(你是谁)与”状态层”(你今天怎么样)的分层,那么把两层分别拿去跨年认人,身份层应该大获全胜、状态层应该一败涂地——而这个对比本身,就是对分层假说最直接的一次检验。

    正文

    上一篇文章我们论证了:四路签名(IAF、低阶谐波、QRS 形态 vs HRV、高阶谐波)里,身份层自己”组内稳、组间散”,状态层相反——但它们从未被放进同一个检验里比过。这一篇换个更锋利的问法:直接让两层去”认人”,看谁认得出、谁认不出。

    为什么识别任务能当试金石?因为”识别”和”信度”是两个不同的概念。信度(ICC)只问”同一个人的读数稳不稳”,识别(AUC)问的是”把 A 和 B 的读数放在一起,机器能不能分清谁是谁”。一个人可以重测很稳(高 ICC),但人人长得一样(识别率≈随机);也可以重测一般稳,但彼此差异足够大(识别率很高)。分层假说的预言是:身份层参数在跨年识别任务里必须赢,状态层参数必须输——否则”分层”就是一张空头支票。

    身份层:跨年认人,文献已经给出了答案

    身份层三条支路各自都有跨会话识别的实测:

    脑电谱签名是最成熟的。2007 年的一项研究让 55 名健康人做静息脑电,其中 20 人隔 12-40 个月重测,用频谱形状做”认人”:40 次会话里 35 次被正确匹配,全部 2960 次两两比较中 99.5% 判对(PMID 17892969)。2021 年天津大学团队把时间拉长到三次实验、每次间隔至少两周,四类特征(功率谱密度、互谱、通道相干、相位滞后)的跨实验识别率仍达 80-100%(PMID 34588964)。2021 年另一项研究在 6 个月内做了 6 次记录,EEG 功能连接的个体特异性识别精度始终高于 90%,六个月后”只发生轻微变化”(PMID 34679331)。脑电侧还有一条”特质”证据:IAF 高度可遗传、跨年龄稳定、100 小时认知训练纹丝不动(PMID 23551082)——这为”身份层为什么稳”提供了生理层解释。

    心电 QRS 形态同样扛得住时间。2018 年一项 460 人的研究,两次 12 导心电记录间隔超过 1 年:单导联(II 导)跨年验证率 86.8%,12 导联拼起来达 97.5%(PMID 29771930)。同一团队的另一项研究把识别任务拆开,间隔一年以上的 12 导 PQRST/QRS 模板,在参考库 10-230 人范围内识别准确率 98.4%-87.4%(PMID 29382064)。心率的即时变化(0 到 >20 bpm)对识别率没有影响——注意,这正是”状态”在变,而”身份”纹丝不动。

    状态层:跨年认人,文献给的答案恰恰相反

    状态层的代表是 HRV。文献给出的画面完全不同:一篇系统性综述直接总结”把 HRV 描述为可靠测量是严重简化”——信度系数高度异质,变异系数可高达 100%(PMID 16098384)。隔日重测的具体数字更扎眼:39 名健康人、严格控制的实验室条件下,第二天同一人的 SDNN 可以是第一天的 0.5 倍到 1.9 倍,LF/HF 最夸张,0.3 倍到 3.5 倍(PMID 17381425)。个体水平日间随机波动这么大,意味着拿 HRV 当”指纹”去跨年认人,基本等于拿天气当身份证。

    缺口:三路对比,从来没人做过

    但注意——以上所有证据都是”各说各话”:脑电说自己稳,心电说自己稳,HRV 说自己不稳。没有人把身份层和状态层放进同一个识别任务里正面比过,更没有人回答三个具体问题:

    1. 只用身份层建模(低阶谐波/IAF/QRS 形态),跨年识别率有多高?
    2. 只用状态层建模(HRV/高阶谐波),跨年识别率掉到多少?
    3. 全参数建模,识别率比只用身份层高还是低?——如果”全参数”比”只用身份层”高,说明状态层里混进了身份信息(污染);如果反而低,说明状态层的噪声在拖后腿。

    第三个问题才是这篇的落点:状态层污染量化。若 HRV 这类”状态参数”在跨年识别里也能贡献识别率,那要么是它携带了不该有的稳定个体信息(如遗传决定的基线水平),要么是模型把个体差异当成了状态差异——无论哪种,既往”状态层=即时状态”的解读都要打折扣。

    建议的裁决设计

    千人队列(覆盖 20-80 岁、健康与慢病),同一协议采集四路签名;身份层特征(IAF、QRS 形态、低阶谐波相位)与状态层特征(HRV 频域、高阶谐波能量比)分别建模,另设全参数组合;训练集与测试集间隔 ≥1 年;以跨年识别率与 AUC 对比,并按训练-测试时间间隔分层(1 年/3 年/5 年)看衰减斜率。若身份层 AUC 显著高于状态层且全参数组合不高于身份层,分层假说获得最强支持;若全参数显著高于身份层,状态层存在污染、须先做特征去污;若状态层 AUC 也高于 0.8,”状态层”这个命名本身就要重审。

    证据与局限

    已验证: IAF 跨年龄稳定、遗传度高(PMID 23551082);静息脑电谱 12-40 个月跨会话识别 99.5% 配对正确(PMID 17892969);EEG 功能连接 6 个月识别精度 >90%(PMID 34679331);跨会话静息脑电识别 80-100%(PMID 34588964);12 导心电间隔 >1 年跨年验证率 97.5%、识别 98.4-87.4%(PMID 29771930、PMID 29382064)。
    部分验证: HRV 隔日重测个体波动极大(SDNN 0.5-1.9 倍、LF/HF 0.3-3.5 倍),信度文献高度异质(PMID 17381425、PMID 16098384)——状态层”认不出人”的方向与分层假说一致,但从未与身份层在同一任务内对比。
    待验证: 三路对比(身份层/状态层/全参数)跨年识别率的正面交锋;状态层污染量化的具体数值;训练-测试间隔对两层识别率衰减斜率的不同影响。

    相关阅读

    L1 网络状态:该主题对应的边(「个体频率签名 · 身份层/状态层识别率对比」)在 CORRESPONDENCE_NETWORK 中不存在——与 #546 同属「个体频率签名」簇(该簇首条缺口已于 2026-10-24 登记为 L1 gap 第 40 条):节点表生理侧仅有 HRV、呼吸率、EEG 等横断面观测项,无「参数分层/识别任务/跨会话稳定性」类测量学节点;边表止于 E-050,无任何以识别率、身份层、状态层为源或目标的边。本文证据等级因此受限于 B。

    证据等级: B
    参考文献: PMID 23551082, PMID 17892969, PMID 34588964, PMID 34679331, PMID 29771930, PMID 29382064, PMID 17381425, PMID 16098384


    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。

    Recognizing People, Not Just Names: A Comparative Analysis of Identity Layer and Status Layer Performance in Year-End Recognition Tasks

    Evidence level: B

    Core Proposition: If frequency signatures genuinely contain a stratification between an “identity layer” (who you are) and a “state layer” (how you are today), then applying each layer separately to cross-year recognition tasks should result in the identity layer winning decisively and the state layer failing completely. This contrast itself constitutes the most direct test of the stratification hypothesis.

    Main Text

    In the previous article, we demonstrated that among the four signature pathways (IAF, low-order harmonics, QRS morphology vs. HRV, high-order harmonics), the identity layer exhibits “stability within groups and dispersion between groups,” while the state layer shows the opposite pattern. However, they have never been subjected to the same comparative test. This article poses a sharper question: directly pit the two layers against each other in “recognition” tasks to see which one succeeds and which one fails.

    Why can recognition tasks serve as the ultimate test? Because “recognition” and “reliability” are distinct concepts. Reliability (ICC) asks only “is the reading stable for the same person?” whereas recognition (AUC) asks “when placing the readings of Person A and Person B together, can the machine distinguish who is who?” A person can have highly stable retest results (high ICC), yet everyone may look identical (recognition rate ≈ random); conversely, retest stability can be moderate, yet differences between individuals may be sufficiently large (high recognition rate). The prediction of the stratification hypothesis is that identity layer parameters must win in cross-year recognition tasks, while state layer parameters must lose; otherwise, “stratification” is merely an empty promise.

    Identity Layer: Cross-Year Recognition of People, Literature Already Provides the Answer

    Three branches of the identity layer each have empirical evidence of cross-session recognition:

    EEG spectral signatures are the most mature. A 2007 study had 55 healthy individuals undergo resting EEG, with 20 retested after 12–40 months. Using spectral shape for “recognition,” 35 of 40 sessions were correctly matched, and 99.5% of all 2,960 pairwise comparisons were correct (PMID 17892969). In 2021, a Tianjin University team extended the timeline to three experiments with intervals of at least two weeks between each; recognition rates for four feature types (power spectral density, cross-spectrum, channel coherence, phase lag) remained between 80–100% (PMID 34588964). Another 2021 study recorded six times over six months; individual-specific recognition accuracy of EEG functional connectivity consistently exceeded 90%, with only “slight changes” after six months (PMID 34679331). On the EEG side, there is also “trait” evidence: IAF is highly heritable, stable across ages, and unaffected by 100 hours of cognitive training (PMID 23551082)—providing a physiological explanation for why the identity layer remains stable.

    ECG QRS morphology also withstands time. A 2018 study of 460 individuals compared two 12-lead ECG recordings separated by over one year: single-lead (Lead II) cross-year verification reached 86.8%, and combining all 12 leads achieved 97.5% (PMID 29771930). Another study by the same team split the recognition task; 12-lead PQRST/QRS templates with intervals over one year achieved recognition accuracy of 98.4%–87.4% within a reference library of 10–230 people (PMID 29382064). Immediate heart rate changes (0 to >20 bpm) had no impact on recognition rates—note, this is precisely where “state” varies while “identity” remains unchanged.

    State Layer: Cross-Year Recognition of People, Literature Provides the Opposite Answer

    The representative of the state layer is HRV. The literature presents a completely different picture: one systematic review directly summarizes that “describing HRV as a reliable measure is a severe oversimplification”—reliability coefficients are highly heterogeneous, with coefficients of variation reaching up to 100% (PMID 16098384). Specific numbers for day-to-day retesting are striking: in 39 healthy individuals under strictly controlled laboratory conditions, the SDNN of one person on Day 2 could be 0.5 to 1.9 times that of Day 1, and the LF/HF ratio ranged from 0.3 to 3.5 times (PMID 17381425). Such large individual-level daily random fluctuations mean using HRV as a “fingerprint” for cross-year recognition is akin to using the weather as an ID card.

    Gap: Three-Way Comparison Never Done

    However, note—all the evidence above is “each talking to itself”: EEG claims stability, ECG claims stability, HRV claims instability. No one has placed the identity layer and state layer in the same recognition task for a head-to-head comparison, nor have they answered three specific questions:

    1. Modeling using only the identity layer (low-order harmonics/IAF/QRS morphology): what is the cross-year recognition rate?
    2. Modeling using only the state layer (HRV/high-order harmonics): how far does the recognition rate drop?
    3. Full-parameter modeling: is the recognition rate higher or lower than using only the identity layer?—If “full parameters” is higher than “identity layer alone,” it indicates the state layer contains identity information (contamination); if instead lower, it means noise in the state layer is dragging performance down.

    The third question is this article’s focus: quantifying state layer contamination. If HRV and other “state parameters” contribute to recognition rates in cross-year identification, either they carry stable individual information that should not be there (e.g., genetically determined baseline levels), or the model mistakes individual differences for state variations. In either case, the traditional interpretation of “state layer = immediate state” must be discounted.

    Proposed Decision Design

    A thousand-person cohort (covering ages 20–80, healthy and chronic disease), collecting all four signature pathways under the same protocol; identity layer features (IAF, QRS morphology, low-order harmonic phase) and state layer features (HRV frequency domain, high-order harmonic energy ratios) modeled separately, plus a full-parameter combination; training and test sets separated by ≥1 year; compare cross-year recognition rates and AUC, stratified by training-test interval (1 year/3 years/5 years) to observe decay slopes. If the identity layer AUC is significantly higher than the state layer and the full-parameter combination does not exceed the identity layer, the stratification hypothesis receives its strongest support; if the full-parameter combination is significantly higher than the identity layer, the state layer contains contamination and feature decontamination must be performed first; if the state layer AUC also exceeds 0.8, the naming of “state layer” itself must be re-examined.

    Evidence and Limitations

    Verified: IAF stability across ages and high heritability (PMID 23551082); cross-session recognition of resting EEG spectra at 12–40 months achieved 99.5% correct pairs (PMID 17892969); EEG functional connectivity recognition accuracy >90% over six months (PMID 34679331); cross-session resting EEG recognition 80–100% (PMID 34588964); cross-year verification rate of 12-lead ECG intervals >1 year at 97.5%, recognition accuracy 98.4–87.4% (PMID 29771930, PMID 29382064).
    Partially Verified: HRV day-to-day retesting shows extreme individual variability (SDNN 0.5–1.9 times, LF/HF 0.3–3.5 times), with reliability literature showing high heterogeneity (PMID 17381425, PMID 16098384)—the direction of the state layer “failing to recognize people” aligns with the stratification hypothesis, but it has never been compared within the same task as the identity layer.
    To Be Verified: head-to-head cross-year recognition rates of the three-way comparison (identity layer/state layer/full parameters); specific numerical quantification of state layer contamination; differential effects of training-test intervals on the decay slopes of recognition rates for both layers.

    Related Reading

    L1 Network Status: The edge corresponding to this topic (“Individual Frequency Signature · Identity Layer/State Layer Recognition Rate Comparison”) does not exist in the CORRESPONDENCE_NETWORK—along with #546, both belong to the “Individual Frequency Signature” cluster (the first gap of this cluster was registered as L1 gap #40 on 2026-10-24): the node table physiological side contains only cross-sectional observations such as HRV, breathing rate, EEG, etc., lacking measurement nodes for “parameter stratification/recognition tasks/cross-session stability”; the edge table stops at E-050, with no edges originating from or targeting recognition rates, identity layer, or state layer. Consequently, this article’s evidence level is limited to B.

    Evidence Level: B
    References: PMID 23551082, PMID 17892969, PMID 34588964, PMID 34679331, PMID 29771930, PMID 29382064, PMID 17381425, PMID 16098384


    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.