Category: 科学易学

  • 丹田参数的「信噪比排序」能不能跨模态复现:X5/AX 的设备偏倚真的最大吗

    Evidence level: C

    排序不是测出来的,是比较出来的;而比较用的那把尺子,本身也有刻度。

    一、信噪比是比值,分母小的时候它不老实

    命题要求把五个参数按「设备偏倚 ÷ 效应量」排个高低。这个量是比值,分子分母都带误差,比值估计量的方差不是两者方差之和——两个估计相关时还要减去协方差项,而分母接近零时它会爆炸。X5 与 AX 恰恰是分母最不稳定的一类:强迫振荡的电抗类参数重复性明显弱于电阻类,学龄前儿童实测 X5 的组内相关仅 0.55、95% 置信区间下限掉到 −0.17,相对重复系数 91.8%,而同批人的 R5 是 0.82 与 35.5%(PMID 35927844)。分母自己就抖成这样,比出来的比值自然也抖。

    二、排序不稳定有实测:单队列的第十名,重抽能进前三

    英国 Trent 癌症登记处做过一次直接的检验:把地区生存率排名反复随机重抽,一个在 1 年乳腺癌生存率上排第十的地区,6% 的重抽里能进前三;同一地区在三个相邻队列里分别排第十、第五、第十(PMID 11429710)。作者的原话值得记住——排名差异很大,可能只反映实际结局上微不足道的差别。五个参数排五个名次,比几十个地区排名更容易翻转。

    三、「谁最大」是选出来的,选出来的极值天生偏大

    「X5/AX 偏倚最大」这句话,是从五个数里挑出最大的那个得来的。凡是先选后报的极值,都带着赢家诅咒:被选中的估计系统性偏高,重测会向均值回落(PMID 41869282 讨论的正是这个机制)。所以即使第一轮真测出 X5 在顶端,第二轮它掉到第二、第三也是预期之内,不构成对命题的反驳,也不构成对命题的证实——单轮排序根本没有检验力。

    四、排序还取决于用哪个指标当裁判

    人工晶体度数公式的横向比较给了一个干净的示范:八个公式在 330 只超长眼轴上比精度,按均方根绝对误差排,Zhu-Lu 最低(0.562 D),Cooke K6 紧随(0.563 D);可换成中位绝对误差,Cooke K6 反而最低(0.271 D),前者的优势消失(PMID 42177933)。同一批数据、同一批公式,换个评价指标冠军就换人。丹田五参数若分别用 ICC、LoA 宽度、相对重复系数当尺子,排出来的序很可能不是同一个。

    五、三个模态共用同一批人,一致可能是假的

    让 RIP、FOT、EIT 各自排一次序,再算排序一致性,看起来很稳。麻烦在于三个模态测的是同一批受试者:体型、年龄、呼吸模式同时影响三处的效应量,会制造出共同的相关结构,让三个排序「自发地」相像。这类假设违背会把一致性系数推高——有研究实测,忽略方差异质性时组内相关被高估到 0.640,忽略研究间变异再抬 0.02–0.07(PMID 30208858)。一致本身不能证明命题,只能证明三个模态被同一批人牵着走。

    六、设备偏倚属于设备对,不属于参数

    这是最硬的一处。两台商用振荡仪对 49 例患者加 27 名健康人的配对测量显示:全部阻抗参数都显著不同,AX 的系统偏倚 −4.10 cmH₂O/L,一致性界限 −14.99 到 6.80,ICC 却还在 0.59–0.85 之间;结论是这两台机器不能互换使用(PMID 42871008)。请注意「这两台」。换一台参考机、换一条呼吸带、换一个电极环,偏倚就换了。所以「X5/AX 偏倚在所有参数中最大」这句话缺一个主语——哪一对设备。

    反转:该报的是双栏表加置信集,不是排序

    把命题改写成可回答的形式:对指定的设备对,逐参数同时报偏倚及其置信区间、效应量及其置信区间,再报排序的 bootstrap 置信集。排序只在设备对层面成立,不能上升为参数属性。

    四格判据:三个模态都排出 X5/AX 在顶端、且相邻名次差的区间不跨零——排序可报,但限定于本设备对;三模态一致而前两名差跨零——只能报「X5/AX 属偏高组」,不许报「最大」;一致性系数显著但换设备对后翻转——排序是设备对属性,禁止上升;一致性系数不显著——退回逐参数原文偏倚与一致性界限,排序这条线直接关闭。

    证据与局限

    已验证:各参数设备偏倚与一致性界限确实差异很大,且电抗类参数的重复性弱于电阻类(PMID 42871008、35927844、22960661);单队列排名不稳定、极值估计偏高、指标替换会改变名次,这三条在别的领域都有实测(PMID 11429710、41869282、42177933)。

    部分验证:排序确定性本身已有量化工具,网络 meta 分析里提出过 POTH 指标与秩聚类方法(PMID 40542622、41635948),小评判者数下的一致性系数区间估计也有专门比较(PMID 24605969);但这些工具都还没用在丹田参数上。

    待验证:X5/AX 是否真的在五参数中偏倚最大——现有文献只能证明某两台机器之间 AX 偏倚大,无法证明它在所有设备对、所有人群里都最大。跨模态排序是否复现,从未被检验过。

    L1 网络状态:该主题尚未进入 L1 对应网络。网络节点表只覆盖易学/中医侧、物理侧、生理/体验侧三类节点,无「设备偏倚 / 信噪比 / 排序一致性」等测量学维度节点;边表当前只到 3.4 天干地支↔五行;「丹田」仅在概念计数中出现(51 篇,最高 A 级 11 篇),未形成对应边。

    相关阅读

    证据等级: C
    参考文献: PMID 42871008, PMID 35927844, PMID 22960661, PMID 30208858, PMID 41869282, PMID 42177933, PMID 11429710, PMID 40542622, PMID 41635948, PMID 24605969


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

    Can the \”Signal-to-Noise Ratio Ranking\” of Dan Tian Parameters Be Reproduced Across Modalities: Is the Device Bias of X5/AX Truly the Largest?

    Evidence level: C

    Ranking is not measured; it is comparative. And the ruler used for comparison also has its own scale.

    I. Signal-to-Noise Ratio is a Ratio; When the Denominator is Small, It Becomes Unstable

    The proposition requires ranking five parameters by “Device Bias ÷ Effect Size.” As a ratio, both numerator and denominator carry error. The variance of the ratio estimator is not merely the sum of the two variances; when estimates are correlated, a covariance term must be subtracted. Furthermore, as the denominator approaches zero, the variance explodes. X5 and AX belong precisely to the category with the most unstable denominators: reactance-type parameters forced into oscillation exhibit significantly weaker repeatability than resistance-type parameters. In pre-school children, the intraclass correlation (ICC) for X5 was only 0.55 with a 95% confidence interval lower bound dropping to −0.17, and the relative repeatability coefficient reached 91.8%. In contrast, R5 from the same cohort showed an ICC of 0.82 and a repeatability coefficient of 35.5% (PMID 35927844). If the denominator itself is so unstable, the resulting ratio will naturally fluctuate.

    II. Ranking Instability Has Empirical Evidence: A Single Queue’s Tenth Place Can Enter the Top Three Upon Resampling

    The UK Trent Cancer Registry conducted a direct test: by repeatedly randomly resampling regional survival rates, a region ranked tenth in one-year breast cancer survival appeared in the top three in 6% of resamples. The same region ranked tenth, fifth, and tenth across three adjacent cohorts (PMID 11429710). The author’s own words are worth remembering: ranking differences can be large, potentially reflecting only trivial differences in actual outcomes. Ranking five parameters against five ranks is far more prone to flipping than ranking dozens of regions.

    III. “Who is Largest” Is Selected; Selected Extremes Are Naturally Biased High

    The statement “X5/AX bias is the largest” is derived by selecting the maximum value from five numbers. Any extreme selected after the fact carries the winner’s curse: the selected estimate is systematically biased upward, and re-measurement will regress toward the mean (PMID 41869282 discusses this mechanism). Therefore, even if X5 truly appears at the top in the first round, dropping to second or third in a second round is expected and does not refute the proposition nor confirm it. Single-round ranking lacks statistical power for validation.

    IV. Ranking Depends on Which Metric Serves as the Judge

    A clean demonstration comes from comparing artificial lens power formulas across eight formulas in 330 ultra-long axial lengths. Ranked by root mean square absolute error, Zhu-Lu was lowest (0.562 D), followed closely by Cooke K6 (0.563 D). However, using median absolute error, Cooke K6 became the lowest (0.271 D), causing Zhu-Lu’s advantage to vanish (PMID 42177933). With the same data and formulas, changing the evaluation metric changes the champion. If the five Dantian parameters are measured using ICC, LoA width, or relative repeatability coefficient as the ruler, the resulting ranking will likely differ.

    V. Three Modalities Share the Same Subjects; Consistency May Be False

    Allowing RIP, FOT, and EIT to each generate a ranking independently, then calculating ranking consistency, appears stable. The complication lies in the fact that three modalities measure the same cohort of subjects: body habitus, age, and breathing patterns simultaneously affect effect sizes across all three modalities, creating a shared correlation structure that causes the three rankings to resemble each other “spontaneously.” Such assumption violations inflate consistency coefficients—studies have measured intraclass correlation overestimated to 0.640 when variance heterogeneity is ignored, and further inflated by 0.02–0.07 when between-study variation is ignored (PMID 30208858). Consistency alone cannot prove the proposition; it only proves that the three modalities are being led by the same cohort.

    VI. Device Bias Belongs to Device Pairs, Not Parameters

    This is the most rigorous point. Paired measurements of 49 patients and 27 healthy subjects using two commercial oscillators showed that all impedance parameters were significantly different. AX exhibited a systematic bias of −4.10 cmH₂O/L with limits of agreement ranging from −14.99 to 6.80, yet the ICC remained between 0.59 and 0.85; the conclusion was that these two machines cannot be used interchangeably (PMID 42871008). Note the phrase “these two machines.” Changing a reference machine, a breathing belt, or an electrode ring changes the bias. Therefore, the statement “X5/AX bias is the largest among all parameters” lacks a subject—specifically, which device pair.

    Reversal: Report Dual-Column Tables with Confidence Sets, Not Rankings

    Reformulate the proposition into a testable form: For a specified device pair, simultaneously report bias and its confidence interval for each parameter, effect size and its confidence interval, and then report the bootstrap confidence set for the ranking. Ranking holds only at the device-pair level and cannot be elevated to a parameter attribute.

    Four-Grid Criterion: If all three modalities rank X5/AX at the top AND the interval for adjacent rank differences does not cross zero—ranking can be reported but must be limited to this specific device pair. If three modalities agree but the interval for the top two ranks crosses zero—only report “X5/AX belongs to the high-bias group”; do not report “largest.” If consistency coefficients are significant but flip upon changing the device pair—ranking is a device-pair attribute; prohibit elevation. If consistency coefficients are not significant—revert to raw parameter bias and limits of agreement; close the ranking line entirely.

    Evidence and Limitations

    Verified: Significant differences exist in device bias and limits of agreement across parameters, with reactance-type parameters showing weaker repeatability than resistance-type (PMID 42871008, 35927844, 22960661); single-cohort ranking instability, winner’s curse in extreme estimates, and metric substitution altering ranks are empirically verified in other fields (PMID 11429710, 41869282, 42177933).

    Partially Verified: Tools for quantifying ranking determinism exist, including the POTH index and rank clustering methods proposed in network meta-analysis (PMID 40542622, 41635948), and specialized comparisons for consistency coefficient interval estimation with small numbers of judges (PMID 24605969); however, none of these tools have yet been applied to Dantian parameters.

    To Be Verified: Whether X5/AX truly exhibits the largest bias among the five parameters—existing literature can only prove AX bias is large between two specific machines, unable to prove it is maximal across all device pairs and populations. Whether cross-modal ranking replicates has never been tested.

    L1 Network Status: This topic has not yet entered the L1 corresponding network. The node table covers only Yijing/TCM, Physics, and Physiological/Experiential nodes; there are no measurement dimensions nodes for “Device Bias / Signal-to-Noise Ratio / Ranking Consistency.” The edge table currently only covers the 3.4 Heavenly Stems and Earthly Branches ↔ Five Phases mapping; “Dantian” appears only in concept counting (51 papers, highest A-level 11), without forming a corresponding edge.

    Related Reading

    Evidence Level: C
    References: PMID 42871008, PMID 35927844, PMID 22960661, PMID 30208858, PMID 41869282, PMID 42177933, PMID 11429710, PMID 40542622, PMID 41635948, PMID 24605969


    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.

  • 丹田呼吸「气感特异分量」的减法陷阱:等频率伪呼吸对照测不出它想测的东西

    Evidence level: C

    想用「等频率伪呼吸」做减法,前提是两条件只差一个成分;可 X5 和 AX 恰恰被这个成分以外的每一种机械变量拉着走。

    正文

    一、减法设计的入场券是「纯插入」

    任何「A 减 B 得到某个成分」的设计,都默认一件事:往任务里加进那个成分时,其余部分一分不动。这个假设叫纯插入,认知科学界为它吵了三十年,结论是它在真实系统里几乎总不成立——两个条件不仅成分多了,还会互相改变(PMID 9345501)。放到呼吸力学上更麻烦:呼吸不是可拆的加工阶段,而是闭环系统,改一处必改另一处。

    二、X5 和 AX 本来就被机械变量牵着走

    已有实测很清楚。胸壁的弹性与阻力随呼吸频率与潮气量一起变,肺本身在平均肺容积不变时才近乎恒定(PMID 1731572);阻力与电抗都随肺容积变化,低于功能残气量时频率依赖性还会自己冒出来(PMID 7204163)。高频段的阻抗同样随肺容积下降(PMID 17943446)。用物理模型直接灌气,呼吸频率从 10 增到 30、潮气量从 0.5 增到 1.5 L,X5 会单调抬高(PMID 25034109)。在 COPD 患者身上把呼吸频率从 20 提到 40,呼气相电抗显著下降,R5−R20 与共振频率同时上升(PMID 25251948)。

    也就是说:等频率只锁住「呼吸频率」一个变量。潮气量、呼气末容积、胸腹贡献比、吸气波形,全都还自由。

    三、等频率伪呼吸并不等容积轨迹

    丹田呼吸的核心特征恰恰是容积轨迹的改动:呼气末位置抬高、腹式占比上升、吸气上升沿变缓。一个「等频率的非丹田式伪呼吸」,频率对上了,这三样未必对上;而这三样每一样都单独能推着 X5 走。于是 ΔX5 = ΔX5_丹田 − ΔX5_伪呼吸 减掉的不是机械分量,而是两个机械条件本就不一致的测量之差。残差里既可能有气感特异成分,也可能全是机械泄漏,单看一个差值无法分辨。

    四、差值自己的信度比两个分量都低

    这是最容易被忽略的一层。两个条件各自的分数往往高度相关,相减时共同的真分差被减掉,两边的误差项却相加,结果差值的信度系统性低于分量(PMID 36842498)。在一批实验任务上的实测正印证了这一点:以差值为主要指标的测量,重测信度排在全部指标的最末尾,而同一个任务里不用差值的单维指标信度良好(PMID 39890691)。

    放到这里意味着:即使 X5 的重测一致性勉强可用,ΔX5 的一致性可能已经掉到无法支撑组间比较的区间。而 FOT 参数的日间波动本就有 5%–15% 的量级,电抗一侧比电阻一侧更差。

    五、期望与盲法这一侧也漏着

    受试知道哪一条件是「真功法」。「深慢呼吸」的心理效应本身可被期望启动放大:同样三分钟深呼吸,带运动期望的那组幸福感提升更多、正性情绪也才显著上升(PMID 27231047)。连生理盐水雾化都能在近半数受试身上缓解实验室诱发的呼吸困难,而把期望明确压到最低后这种效应就少见(PMID 28004181)。

    所以「伪呼吸」这个名字不构成对照。真要检验,必须把盲法本身测出来:让受试猜分配并标信心,算盲法指数。某大型假手术试验里,真治疗臂的指数 0.397、假治疗臂 −0.396,两侧大致对称——这叫「一厢情愿」,不是成功盲法(PMID 31673601)。

    六、AX 还有一个独立的结构性问题

    AX 是从 5 Hz 积分到共振频率的电抗面积。共振频率测不出来时,积分上限由使用者自定,不同厂商的设备起始频率也不一样(成人 4 Hz、儿童可到 8 Hz),所以 AX 的绝对值在设备之间不可直接互换;而且 AX 与个体间通气缺陷比例只有弱相关,多变量模型里根本不显著(PMID 30632309)。用它作「特异增量」的载体,等于把判据建在一个半自定义的量上。

    证据与局限

    已验证: 呼吸频率、潮气量、肺容积单独都能显著改变 X5 与阻力频率依赖性(PMID 1731572、PMID 7204163、PMID 17943446、PMID 25034109、PMID 25251948);以差值为指标的测量其重测信度系统性低于其分量(PMID 36842498、PMID 39890691);期望可独立于干预本身改变呼吸相关的主观体验(PMID 27231047、PMID 28004181)。

    部分验证: 「纯插入」假设在认知神经科学中被明确判定为多数场合不成立(PMID 9345501),但尚无研究在呼吸力学里定量标定「等频率不等于等机械状态」的偏差幅度;AX 的设备依赖性与非特异性有明确文献(PMID 30632309、PMID 35140105、PMID 33033141),但「气感特异增量」这一构念本身从未被任何研究测量过。

    待验证: ①把伪呼吸调到与丹田呼吸在潮气量、呼吸频率、呼气末容积、腹式占比四个维度逐点匹配后,匹配残差能否压到足以让残余 ΔX5 可解释;②ΔX5 与 ΔAX 自身的重测一致性是否达到可作终点的水平;③气感评分与残余 ΔX5 的关联在盲法成立时是否还在。

    本研究的四格判据:四维匹配残差均低于预设阈值、残余 ΔX5 仍显著、且 ΔX5 的组内相关 ≥0.75 ⇒ 特异增量成立;匹配残差大且残余 ΔX5 随残差单调缩放 ⇒ 全是机械泄漏、所谓特异分量不存在;匹配良好但残余不显著 ⇒ 气感与 X5/AX 无独立关联,原假说方向要改;ΔX5 重测一致性低于 0.5 ⇒ 差值终点不可用,须退回单条件水平比较或改主观-客观配对设计。

    一条改判:真正该报的不是「气感特异增量」,而是匹配残差表 + 残余差值信度。先把「伪呼吸到底等不等」写成可核对的数字,再谈那一块残余归谁。

    相关阅读

    L1 网络状态:该主题尚未进入 L1 对应网络。CORRESPONDENCE_NETWORK.md 节点表中「呼吸率 / 呼吸模式」仅作生理侧观测项、「气感报告」仅作结构化自评项,两者均未成边;边表只列到 3.4「天干地支 ↔ 五行」,无「呼吸力学分量分离 / 减法设计 / 差值信度」维度。故本文证据等级封顶 C。
    证据等级: C
    参考文献: PMID 9345501, PMID 1731572, PMID 7204163, PMID 17943446, PMID 25034109, PMID 25251948, PMID 36842498, PMID 39890691, PMID 27231047, PMID 28004181, PMID 31673601, PMID 30632309, PMID 35140105, PMID 33033141


    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 Subtraction Trap of \”Qi Sensation Specificity\” in Dan Tian Breathing: Frequency-Matched Pseudo-Breathing Fails to Reveal What It Aims to Measure

    Evidence level: C

    Using “equal-frequency pseudo-breath” for subtraction presupposes that only one component differs between conditions; however, X5 and AX are both dragged by every mechanical variable outside that single component.

    Main Text

    I. The Ticket for Subtraction Design is “Pure Insertion”

    Any design assuming “A minus B yields a specific component” defaults to one thing: adding that component into the task leaves the rest completely unchanged. This assumption is called pure insertion. Cognitive science has debated it for three decades, concluding that it rarely holds in real systems—conditions differ not just by component addition but also by mutual alteration (PMID 9345501). This becomes more problematic in respiratory mechanics: breathing is not a decomposable processing stage but a closed-loop system; changing one part inevitably alters another.

    II. X5 and AX Are Already Being Driven by Mechanical Variables

    Existing empirical data are clear. Chest wall elasticity and resistance vary with breathing frequency and tidal volume, while the lung itself remains nearly constant only when average lung volume is unchanged (PMID 1731572). Resistance and reactance both change with lung volume, and frequency dependence can emerge spontaneously below functional residual capacity (PMID 7204163). Impedance in the high-frequency range also decreases with falling lung volume (PMID 17943446). Using physical models to directly inflate the lungs, increasing breathing frequency from 10 to 30 and tidal volume from 0.5 to 1.5 L causes X5 to rise monotonically (PMID 25034109). In COPD patients, raising breathing frequency from 20 to 40 causes a significant drop in expiratory phase reactance, with R5−R20 and resonant frequency rising simultaneously (PMID 25251948).

    In short: equal frequency locks only the “breathing frequency” variable. Tidal volume, end-expiratory volume, thoraco-abdominal contribution ratio, and inspiratory waveform remain free.

    III. Equal-Frequency Pseudo-Breath Is Not Equal Volume Trajectory

    The core characteristic of abdominal breathing is precisely a modification of the volume trajectory: elevated end-expiratory position, increased abdominal contribution, and a slower inspiratory rise. An “equal-frequency non-abdominal pseudo-breath” matches the frequency but may not match these three; and each of these three alone can drive X5. Thus, $\Delta X5 = \Delta X5_{\text{Dantian}} – \Delta X5_{\text{Pseudo}}$ subtracts not a mechanical component but the difference between two mechanically inconsistent conditions. The residual may contain the qi-sensation specific component or be entirely mechanical leakage; a single difference value cannot distinguish between them.

    IV. The Reliability of the Difference Is Lower Than Either Component

    This is the most overlooked layer. Scores for each condition are often highly correlated; when subtracted, the common true variance is removed while error terms from both sides add up, resulting in systematically lower reliability for the difference than for the components (PMID 36842498). Empirical data from a batch of experimental tasks confirms this: measurements using difference as the primary index rank at the very end for retest reliability, whereas single-dimensional indices without differences within the same task show good reliability (PMID 39890691).

    This implies here: even if X5’s retest consistency is marginally usable, the consistency of $\Delta X5$ may have already dropped below the range capable of supporting between-group comparisons. Meanwhile, FOT parameters exhibit daily fluctuations on the order of 5%–15%, with reactance showing worse stability than resistance.

    V. Expectations and Blinding Are Also Leaking on This Side

    Subjects know which condition is the “true practice.” The psychological effects of “deep slow breathing” can be triggered and amplified by expectation: the group with exercise expectations showed greater happiness improvement and only significant increases in positive affect after three minutes of deep breathing compared to those without (PMID 27231047). Even saline nebulization can relieve lab-induced dyspnea in nearly half of subjects, but this effect becomes rare when expectation is explicitly minimized (PMID 28004181).

    Therefore, the name “pseudo-breath” does not constitute a valid control. To truly test it, one must measure the blinding itself: have subjects guess the allocation and rate confidence, calculating a blinding index. In a large sham surgery trial, the true treatment arm had an index of 0.397 and the sham arm −0.396, roughly symmetric—this is called “wishful thinking,” not successful blinding (PMID 31673601).

    VI. AX Has an Additional Independent Structural Issue

    AX is the reactance area integrated from 5 Hz to the resonant frequency. When resonant frequency cannot be measured, the integration upper limit is self-defined by the user; different manufacturers’ devices start at different frequencies (4 Hz for adults, up to 8 Hz for children), so AX’s absolute value cannot be directly exchanged between devices; moreover, AX shows only weak correlation with inter-individual ventilation defect ratios and is not significant in multivariate models (PMID 30632309). Using it as a carrier for “specific increment” is equivalent to building the criterion on a semi-customized quantity.

    Evidence and Limitations

    Verified: Breathing frequency, tidal volume, and lung volume alone can all significantly alter X5 and resistance frequency dependence (PMID 1731572, PMID 7204163, PMID 17943446, PMID 25034109, PMID 25251948); measurements using difference as the index show systematically lower retest reliability than their components (PMID 36842498, PMID 39890691); expectation can independently alter breath-related subjective experiences separate from the intervention itself (PMID 27231047, PMID 28004181).

    Partially Verified: The “pure insertion” assumption has been explicitly judged invalid in most contexts within cognitive neuroscience (PMID 9345501), but no study has quantitatively calibrated the magnitude of the “equal frequency does not equal equal mechanical state” bias in respiratory mechanics; AX’s device dependence and non-specificity have clear literature support (PMID 30632309, PMID 35140105, PMID 33033141), but the construct of “qi-sensation specific increment” itself has never been measured by any study.

    To Be Verified: ① Can matching residuals be reduced to a level where residual $\Delta X5$ becomes interpretable after adjusting pseudo-breath to match abdominal breathing point-by-point across tidal volume, breathing frequency, end-expiratory volume, and abdominal contribution ratio? ② Do the retest consistencies of $\Delta X5$ and $\Delta AX$ reach levels suitable for endpoints? ③ Does the association between qi-sensation scores and residual $\Delta X5$ persist when blinding is established.

    Our Study’s Four-Grid Criterion: All four-dimensional matching residuals below preset threshold, residual $\Delta X5$ still significant, AND within-subject correlation of $\Delta X5 \geq 0.75$ ⇒ Specific increment established; Large matching residuals AND residual $\Delta X5$ monotonically scales with residuals ⇒ All mechanical leakage, so-called specific component does not exist; Good matching but residual not significant ⇒ No independent association between qi-sensation and X5/AX, original hypothesis direction needs revision; $\Delta X5$ retest consistency below 0.5 ⇒ Difference endpoint unusable, must revert to single-condition level comparison or adopt subject-objective paired design.

    One Re-judgment: What should truly be reported is not “qi-sensation specific increment,” but rather matching residual table + reliability of residual difference. First write down whether “pseudo-breath is truly equal” as verifiable numbers, then discuss who the residual belongs to.

    Related Reading

    L1 Network Status: This topic has not yet entered the L1 corresponding network. The CORRESPONDENCE_NETWORK.md node table lists “breathing rate / breathing pattern” only as physiological observation items and “qi-sensation reports” only as structured self-assessment items; neither has formed an edge. Edge tables only list up to 3.4 “Heavenly Stems and Earthly Branches ↔ Five Phases,” with no dimensions for “respiratory mechanics component separation / subtraction design / difference reliability.” Therefore, this study’s evidence level is capped at C.
    Evidence Level: C
    References: PMID 9345501, PMID 1731572, PMID 7204163, PMID 17943446, PMID 25034109, PMID 25251948, PMID 36842498, PMID 39890691, PMID 27231047, PMID 28004181, PMID 31673601, PMID 30632309, PMID 35140105, PMID 33033141


    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.

  • 「越练越不容易塌」这句话,横断面证不了:DR1/DR2 拐点位置随修炼年限的剂量-反应,先要过年龄共线这一关

    Evidence level: C

    「剂量-反应」这四个字要求的是同一个人练得越久、拐点越往下走;而横断面上的「年限梯度」,回答的其实是另一个问题。

    正文

    一、拐点位置本来就被年龄单独推着走

    Xrs-容积曲线上的 DR1/DR2,是呼气过程里去复张拐点的位置(Nilsen 等,PMID 30702975)。这项研究在健康人里报告 DR1 约 46.1%TLC、DR2 约 37.69%TLC,哮喘组则分别上移到 76.58% 与 56.79%;更关键的是,健康对照组内部两个拐点都随年龄上移,r² 分别 0.68 与 0.71。也就是说,不做任何干预,光长岁数,拐点就自己往上爬。

    其它通道同向。80 名 20-86 岁正常人:共振频率随年龄升高(p<0.003)、平均电抗下降(p<0.004)(PMID 19936180);KORA 397 名肺功能正常者中,R5-R20、AX 与 Fres 两性均随年龄变化(PMID 23691036);ECOPD 1859 人里年龄每增 10 岁,X5 变化 −0.02、AX +0.30、Fres +1.75(PMID 36058907)。

    第一道坎:年限与年龄几乎共线——练二十年的人也长了二十岁,而年龄这一侧的方向恰好与「越练越不塌」相反。

    二、横断面的年限梯度,不等于纵向变化

    有人会说把年龄当协变量扣掉就行。问题是这个动作预设年龄效应可以线性外推,而上面三个 r² 说明它不是。

    更硬的一条:横断面估出的年龄效应,与真正的纵向变化不是一个数。Van Pelt 等在荷兰随访 12 年、4395 名成人、最多五轮后判定,最老年组实测通气功能明显低于按年轻组纵向变化外推的预期,且不是失访造成的(PMID 8173762)。Nakadate 对 326 名男性五年四次测量发现,除 CVmax25 外纵向年降幅都显著小于横断面估计(PMID 1765546);同一作者 1997 年 243 人七年随访再复现,并指出跨断面方程用于时序评价会产生偏倚(PMID 9396252)。丹麦 11288 名健康不吸烟者的新参考值也证明队列效应实存——多数亚组肺功能水平在过去二十年间是上升的(PMID 22583805)。

    一个横断面上「练 15 年组拐点比新手低 5%TLC」的数字,可能全部来自出生队列差异。

    三、自变量「年限」自己也测不准

    剂量-反应的 X 轴是累积练习量,而这根轴在气功研究里几乎都是自报年限。测量误差对斜率的影响不只是「变小」:单变量线性模型里它确实把斜率往零压(回归稀释比实测 0.51,真斜率约为观测值的两倍,PMID 17148467),但模型一旦非线性或多变量,偏差方向可以两边走——低剂量端被高估、高剂量端被低估,一条本无阈值的曲线也能被误差抹得又平又直。

    「曲线变平」有可能不是适应,而是年限误差的产物。更糟的是三重共线:年限与年龄共线、与「练得对不对」共线、与「身体底子好所以练得下去」共线。

    四、「更平」有三个来源,只有一个是你想说的

    一是真适应:呼吸模式改变使呼气末容积更靠上,拐点相对下移。二是天花板:剂量-反应在多数训练领域都非线性饱和,阿尔茨海默病执行功能的贝叶斯剂量-反应模型峰值约在 1000 MET-min/周,两侧均下降(PMID 40882468)。三是选择效应,也最容易被忽略:练得下去的人本来就更好。

    丹麦 1975-78 年出生队列给出量级——按 10 年累积职业暴露看,4 年暴露后腰背疾患风险升 31%,而忽略暴露史的横断面模型只给出它的四分之一(PMID 38530478)。职业流行病学还有机制描述:离岗者死亡率更高,而低暴露组离岗人时占比更大,就业状态于是成为暴露-反应趋势的负混杂,会把真实趋势抹平甚至反转(PMID 8546122)。

    对应到气功:练不下去的人早就退出了。横断面上「年限越长曲线越平」,完全可能是「曲线平的人留下来了」。

    五、现有气功横断面,全都踩在这个坑上

    3 个月、1-3 年、无经验三组各 16 人的横断面研究显示,长期组腘绳肌与腓肠肌反射潜伏期更短、倾斜板平衡时间更长(PMID 16401443)。22 名太极拳练习者与 18 名匹配对照的横断面研究发现练习年限越长、中央前回与楔前叶 VMHC 越低(PMID 32109436)。两篇都用年限-反应写法,都是横断面,都没处理年龄与自选择。

    结论不一定错,但在拿到纵向数据前,这些年限梯度只能算候选信号。

    六、拐点位置这个因变量,自己也未必测得稳

    现行强迫振荡质控盯的是最低频率 R5 的组内变异系数,Hantos 等直接指出这套标准忽略电抗、不足以代表可重复性,建议改用同时涵盖 R 与 X 的变异度量(PMID 37377657)。700 人群体样本显示单次与三次均值差异很小(PMID 33914403),但那是组水平均值差,不是个体拐点位置的重测一致性。DR2 是从曲线拟合出来的一个点,不是刻度。

    反转:该改的不是「拐点位置」,是「斜率」

    叠起来最要命的结论是:「拐点位置随年限下移」这个命题,横断面在设计上就回答不了,纵向又可能估不准。

    真正可做的替代终点是曲线的形态参数——不是拐点在哪,而是呼气末那一段的斜率有多陡。斜率由大量点共同决定,比单个拟合点稳,也不依赖「拐点」这个必须先成立的概念。若非要保留拐点,必须报出它的重测一致性,并证明该一致性窄于预期组间差。

    证据与局限

    已验证:DR1/DR2 是真实可测的曲线特征,且随年龄上移(PMID 30702975);强迫振荡参数的电抗类指标随年龄单调变化(PMID 19936180、PMID 23691036、PMID 36058907);横断面与纵向估出的年龄效应存在系统性差异(PMID 8173762、PMID 1765546、PMID 9396252);健康工作者/幸存者效应可把暴露-反应趋势大幅压平(PMID 38530478、PMID 8546122);自报暴露的测量误差使观测斜率约为真值一半(PMID 17148467)。

    部分验证:气功/太极的年限-反应在横断面上确有信号(PMID 16401443、PMID 32109436),但样本量小、无年龄匹配的纵向验证,不能作为剂量-反应证据。

    待验证:① 修炼者的 DR2 重测一致性(ICC/MDC95)从未被测量;② 曲线斜率是否比拐点位置更稳定,尚无头对头比较;③ 长期练习组的退出率与入组健康门槛从未被报告,选择效应量级未知。

    相关阅读

    证据等级: C
    参考文献: PMID 30702975, PMID 19936180, PMID 23691036, PMID 36058907, PMID 8173762, PMID 1765546, PMID 9396252, PMID 22583805, PMID 17148467, PMID 40882468, PMID 38530478, PMID 8546122, PMID 16401443, PMID 32109436, PMID 37377657, PMID 33914403
    L1 网络状态:该主题尚未进入 L1 对应网络。CORRESPONDENCE_NETWORK.md 节点表生理侧仅有「呼吸率 / 呼吸模式」观测项(备注「与节律耦合最直接」)与「胃肠蠕动=气沉丹田候选物」,物理侧仅有「相位 / 节律」,均未与易学侧节点连边;无「呼气末肺容积 / 气道闭合 / 肺力学曲线拐点 / 训练年限剂量-反应」这类维度节点。EDGE_TABLE_science-yixue.md 仅有「丹田 51 篇 / 气功 48 篇 / 呼吸 189 篇」文章计数,边表只到 3.4 天干地支↔五行。故本文证据等级按硬约束封顶为 C。


    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。

    Cross-sectional data cannot prove \”the more you practice, the harder it is to collapse\”: the DR1/DR2 inflection point depends on a dose-response relationship with training durat…

    Evidence level: C

    The phrase “dose-response” requires that, for the same individual, the inflection point shifts downward with prolonged practice; however, a cross-sectional “duration gradient” addresses a fundamentally different question.

    Main Text

    I. The Inflection Point Is Already Driven Upward by Age Alone

    The DR1/DR2 on the Xrs-volume curve represent the re-inflation inflection points during exhalation (Nilsen et al., PMID 30702975). This study reported DR1 at approximately 46.1% TLC and DR2 at 37.69% TLC in healthy individuals, while shifting upward to 76.58% and 56.79%, respectively, in the asthma group. More critically, both inflection points within the healthy control group shifted upward with age alone (r² = 0.68 and 0.71), without any intervention. Simply aging causes the inflection point to rise.

    Other channels follow the same trend. In a cohort of 80 healthy adults aged 20–86, resonant frequency increased with age (p<0.003) while average reactance decreased (p<0.004) (PMID 19936180). Among 397 lung-function-normal individuals in KORA, R5-R20, AX, and Fres varied with age in both sexes (PMID 23691036). In the ECOPD cohort of 1859 individuals, for every decade increase in age, X5 changed by −0.02, AX by +0.30, and Fres by +1.75 (PMID 36058907).

    First Barrier: Duration and age are nearly collinear—the person practicing for twenty years is also twenty years older, and the direction of this age-related shift opposes the “the harder it is to collapse with more practice” hypothesis.

    II. Cross-Sectional Duration Gradients Do Not Equal Longitudinal Change

    One might argue that age can be treated as a covariate. The problem is this assumes age effects can be linearly extrapolated, which the aforementioned r² values suggest they cannot.

    A stronger point: Age effects estimated cross-sectionally do not equal true longitudinal change. Van Pelt et al., following 4,395 adults in the Netherlands for 12 years with up to five rounds, found that the oldest group’s actual ventilatory function was significantly lower than expected based on longitudinal extrapolation from younger groups, and this discrepancy was not due to loss to follow-up (PMID 8173762). Nakadate found in 326 men across four measurements over five years that, except for CVmax25, longitudinal annual declines were significantly smaller than cross-sectional estimates (PMID 1765546); the same author reproduced this in a seven-year follow-up of 243 people in 1997, noting that cross-sectional equations applied to temporal evaluation introduce bias (PMID 9396252). New reference values for 11,288 healthy non-smokers in Denmark also confirm cohort effects exist—lung function levels for most subgroups were rising over the past two decades (PMID 22583805).

    A figure stating “the inflection point for the 15-year group is 5% TLC lower than novices” on a cross-sectional plane may derive entirely from birth cohort differences.

    III. The Independent Variable “Duration” Itself Is Measured Inaccurately

    The X-axis of the dose-response relationship is cumulative practice volume, yet in Qigong research, this is almost always self-reported duration. Measurement error affects the slope not just by “shrinking” it: in univariate linear models, it indeed compresses the slope toward zero (regression dilution ratio observed at 0.51, true slope approximately twice the observed value, PMID 17148467). However, once models become non-linear or multivariate, the bias direction can shift both ways—low-dose ends are overestimated, high-dose ends underestimated, rendering a curve with no threshold appear flat and straight due to error.

    “Flattening of the curve” may not reflect adaptation but rather a product of duration error. Worse still is triple collinearity: duration collinear with age, collinear with “correct practice,” and collinear with “better physiological baseline enabling sustained practice.”

    IV. Three Sources for “Flatter,” Only One Is What You Intend

    First, true adaptation: altered breathing patterns shift end-expiratory volume higher, moving the inflection point relatively downward. Second, ceiling effects: dose-response relationships in most training domains exhibit non-linear saturation; the Bayesian dose-response model for executive function in Alzheimer’s disease peaks at approximately 1000 MET-min/week, declining on both sides (PMID 40882468). Third, selection effects, often overlooked: those who can sustain practice are inherently better.

    The Danish cohort born 1975–1978 provides the magnitude—based on cumulative occupational exposure over 10 years, lumbar and back disorder risk rose by 31% after 4 years of exposure, whereas cross-sectional models ignoring exposure history yielded only a quarter of that value (PMID 38530478). Occupational epidemiology also describes mechanisms: mortality is higher among those leaving employment, while the proportion of leavers in low-exposure groups is larger; thus, employment status acts as a negative confounder in exposure-response trends, flattening or even reversing the true trend (PMID 8546122).

    Applied to Qigong: those who cannot sustain practice have already withdrawn. The cross-sectional finding that “longer duration correlates with a flatter curve” may simply reflect that “those with flatter curves remain.”

    V. Existing Cross-Sectional Qigong Studies All Fall Into This Trap

    Cross-sectional studies with three groups (3 months, 1–3 years, no experience) of 16 participants each found that the long-term group had shorter reflex latencies for the hamstrings and gastrocnemius and longer tilt-board balance times (PMID 16401443). A cross-sectional study comparing 22 Tai Chi practitioners with 18 matched controls found that longer practice duration correlated with lower VMHC in the precentral gyrus and precuneus (PMID 32109436). Both used duration-response formulations, were cross-sectional, and failed to address age and self-selection.

    The conclusions may not be wrong, but before obtaining longitudinal data, these duration gradients can only be considered candidate signals.

    VI. This Dependent Variable May Also Be Unstable in Practice

    Current forced oscillation quality control focuses on the within-subject coefficient of variation for the lowest frequency R5. Hantos et al. directly pointed out that this standard ignores reactance and is insufficient to represent reproducibility, suggesting instead a variability metric covering both R and X simultaneously (PMID 37377657). While a population sample of 700 showed minimal differences between single and triple mean values (PMID 33914403), this represents a group-level mean difference, not the retest consistency of an individual’s inflection point position. DR2 is a point derived from curve fitting, not a fixed scale mark.

    Reversal: The Issue Is Not “Inflection Point Position,” But “Slope”

    The most critical conclusion when stacked is: the proposition that “inflection point position shifts downward with duration” cannot be answered by cross-sectional design, and longitudinal data may still estimate it inaccurately.

    The viable alternative endpoint is the curve’s morphological parameters—not where the inflection point lies, but how steep the slope is in that end-expiratory segment. Slope is determined by numerous points jointly, making it more stable than a single fitted point and not reliant on the concept of an “inflection point,” which must first be established. If the inflection point must be retained, its retest consistency must be reported, and proven to be narrower than expected between-group differences.

    Evidence and Limitations

    Verified: DR1/DR2 are real, measurable curve features that shift upward with age (PMID 30702975); reactance-based forced oscillation parameters change monotonically with age (PMID 19936180, PMID 23691036, PMID 36058907); systematic differences exist between cross-sectional and longitudinal estimates of age effects (PMID 8173762, PMID 1765546, PMID 9396252); healthy worker/survivor effects can drastically flatten exposure-response trends (PMID 38530478, PMID 8546122); self-reported exposure measurement error reduces observed slopes to approximately half the true value (PMID 17148467).

    Partially Verified: Duration-response relationships in Qigong/Tai Chi show signals in cross-sectional data (PMID 16401443, PMID 32109436), but with small sample sizes and no age-matched longitudinal verification, they cannot serve as dose-response evidence.

    To Be Verified: ① Retest consistency (ICC/MDC95) of practitioners’ DR2 has never been measured; ② Whether curve slope is more stable than inflection point position lacks head-to-head comparison; ③ Attrition rates in long-term practice groups and health thresholds for entry have never been reported, leaving the magnitude of selection effects unknown.

    Related Reading

    Evidence Level: C
    References: PMID 30702975, PMID 19936180, PMID 23691036, PMID 36058907, PMID 8173762, PMID 1765546, PMID 9396252, PMID 22583805, PMID 17148467, PMID 40882468, PMID 38530478, PMID 8546122, PMID 16401443, PMID 32109436, PMID 37377657, PMID 33914403
    L1 Network Status: This topic has not yet entered the L1 corresponding network. The CORRESPONDENCE_NETWORK.md node table lists physiological side observations only as “Respiration Rate / Respiration Pattern” (noted “most directly coupled with rhythm”) and “Gastrointestinal Peristalsis = Candidate for Qi Settling in Dan Tian”; physical side lists only “Phase / Rhythm”; none connect to Yijing side nodes; no dimensional nodes exist for “End-Expiratory Lung Volume / Airway Closure / Lung Mechanics Curve Inflection / Training Duration Dose-Response”. EDGE_TABLE_science-yixue.md contains only article counts (“Dan Tian 51 / Qigong 48 / Respiration 189”), and edge tables only reach the 3.4 Heavenly Stems and Earthly Branches ↔ Five Phases level. Therefore, per hard constraints, this article’s evidence level is capped at C.


    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.

  • 「体位占比降了」不等于「练得更稳」:方差分量结构的组间比较有一个比率的坑

    Evidence level: C

    把总方差拆成「体位、带位、操作员、个体、残差」之后,最诱人的一步是拿五个占比去比修炼者与新手。这一步几乎肯定会出错,因为占比是比率:分子降、分母升,都能让它降。

    正文

    一、占比是比率,不是量

    五个变异源占比加起来必须等于 1。这个约束带来两个后果:任何两个占比天然负相关;「体位占比从 40% 降到 25%」本身没有信息,因为体位方差可能降了 37%,也可能一分没动、只是个体间方差涨了 60%。前者是”人对姿势不敏感了”,后者是”人跟人差得更远了”,方向完全相反,却给出同一个数字。

    成分数据在统计上早有定论:算术均值给不出单纯形内的结果,须先做对数比变换(PMID 38806627);份额数据里绝对量上升也可能表现为份额下降(PMID 28882426)。所以要报的是绝对方差分量,不是占比。

    二、占比不能直接做组间推断

    占比的抽样分布依赖五个分量共享的分母,五个都不独立。对占比直接跑 t 检验,等于假装分母是常数。临床试验里方差齐性检验本身有专门方法——Jackknife、Cochran 一类在重尾与偏态数据下更稳(PMID 37143826);混合模型里若方差随组别变化,正确做法是直接给方差建模(location-scale 型扩展),而非只检验固定效应(PMID 32666394)。

    「两组的方差结构不同」是关于方差函数的假设,得在方差那一层测,不能靠占比描述性对比。

    三、体位效应本身就不是人人都有

    健康人可能根本没有体位效应可降。23 例重度 COPD 患者坐位出现胸廓异步、仰卧消失,上胸廓与腹部坐位同步而仰卧强异步;12 例年龄匹配健康对照,两个体位都没有任何区间异步(PMID 23412901)。

    如果健康修炼者与新手的体位方差都接近零,”修炼者体位占比更低”就是分母造出来的假象。这是典型天花板效应:指标在目标人群没有变异空间,组间比较毫无意义。设计前必须先确认该人群的体位方差显著大于零。

    四、这么小的设计,方差分量本身估不准

    体位只有 2 个水平,操作员往往 1 到 2 个。测量次数少、随访短、残差大方差,都会扩大”非正则”参数区域,使负方差与边界估计变成常态而非异常(PMID 38695394);不同估计量的负估计概率差别很大(PMID 26843971)。

    推断更麻烦:参数边界上似然比、Wald、Score 的渐近分布失效,须改置换法或贝叶斯法(PMID 31460683、PMID 16474914)。加上评判者数量少本身就是 ICC 偏倚的主因(PMID 33818118),三个模态的置信区间很可能宽到跨零——”有差异”和”无差异”都测不出来。

    五、训练降的是哪种「稳」

    技能研究确实支持”专长的人更稳”,但要看清是哪种稳。9 名外科医生各测 4 次:初级住院医分数标准差 5.8(满分 30),高级住院医与主治都是 1.8(PMID 31517685);模拟器训练中练习变异每降 1%,后测升 3.8 分、活体迁移升 3.0 分(PMID 30898374)。

    但这些都是同一个人的重复测量,属交叉设计里的”残差 + 交互”,不是”体位”这个设计因子。手枪射击的对照更说明问题:专家 COP 位移更小、多尺度熵更低,可两组上身-枪相对相位都接近同相,真正差别在结构——专家是枪滞后于身,新手是枪领先于身(PMID 28628398)。稳不等于每个维度都变小,而是结构换了位置。

    六、组间比较的根本麻烦:不是随机分组

    修炼者与新手不是随机分配的,是自选择。能练十年的人,本来就更坐得住、呼吸控制更好——横断面设计分不清”练出来的”和”能练下去的”。志愿者偏差是系统性的:UK Biobank 参与者比一般人群更健康、社会经济状况更好(PMID 28641372)。

    另有一个读法错误须提防。SD-OCT 测儿童与成人:两组视网膜厚度均值没有差异,但儿童 COV 是成人的 3 到 4 倍、ICC 显著更低(0.73–0.99 vs 0.986–0.993,PMID 26808961)。同样的均值、差很多的精度——这种”组间精度差异”真实存在,却最容易被误写成”这组人能力更强”。

    反转:该报的不是占比,是两个绝对量

    整套设计的错在于拿一个比率的分子占比当质量指标。真正该报的是:① 组内信度——用 G 理论的 D 研究算各自的条件信度,看修炼组的 D 系数是否真的更高;② 残差方差的绝对值——绝对噪声到底降了多少。

    这里还有个已知陷阱:G 理论经典版假设残差方差齐性,实际数据常常违背,不建模异方差会把测量质量的差异整个掩盖掉,须按条件算 G 与 D(PMID 41549064)。

    所以,若修炼组的 D 系数没升、只是体位占比降了,那说明变异只是从”体位”搬到了”人身上”——练十年的人彼此差得更远,这完全可预期,与”测量更稳”毫无关系。

    四条判法:一组体位绝对方差显著更低、D 系数更高、残差绝对方差也更低,三处同向 ⇒ 真稳健性提升;体位绝对方差没降但占比降了 ⇒ 比率假象(分母变大),不可报”训练降低体位敏感”;残差方差反而更大而 D 系数不变 ⇒ 变异只是搬了家,是”换了一种不稳”;两组体位绝对方差都接近零 ⇒ 天花板效应,该指标在此人群不可用,须换带位或操作员。

    配套三条:体位×组、带位×组、操作员×组三个交互项都要报,因为组间方差差异可能只是某因子对某组根本不起作用;报方差分量必须同时报置信区间与估计方法,落在边界上的分量不许报 p<0.05;样本量按”要分辨的方差比”反解,不按均值差反解。

    证据与局限

    已验证: 成分数据的闭合约束导致伪相关、份额算术均值可越界(PMID 28882426);对数比变换处理份额数据的必要性(PMID 38806627);方差齐性检验的多种方法及其在重尾/偏态数据下的表现差异(PMID 37143826);混合模型异方差下须对异方差本身建模而非仅检验固定效应(PMID 32666394);小测量次数与短随访扩大负方差/边界区域(PMID 38695394);不同方差分量估计量的负估计概率差异(PMID 26843971);参数边界上经典检验失效、须置换或贝叶斯法(PMID 31460683、PMID 16474914);评判者少导致 ICC 偏倚(PMID 33818118);G 理论需扩展以处理异方差(PMID 41549064);COPD 与健康对照在体位上的胸廓异步差异(PMID 23412901);外科住院医个体内变异随年资下降(PMID 31517685);模拟器练习变异与技能迁移的关联(PMID 30898374);手枪射击专家与新手的协调结构差异(PMID 28628398);UK Biobank 参与者的健康志愿者偏差(PMID 28641372);儿童与成人 OCT 均值相同而精度不同(PMID 26808961)。

    部分验证: 上述技能变异证据来自外科模拟训练与射击任务,与丹田呼吸的”训练”在任务结构、反馈方式、评估对象上均不同构,只能作方向性参照;OCT 的组间精度差异是仪器测量场景,与”人主动控制呼吸”的机制不同;体位效应的组间异质性证据来自 COPD 与健康对照,未涵盖长期呼吸训练人群。

    待验证: 丹田呼吸场景下体位绝对方差在修炼者与新手间是否真有差异(无任何直接研究);修炼者的 G/D 系数是否真的更高;”修炼者主动固定姿势”造成的方差压缩量级;体位×组的交互是否存在;五个方差分量在 2 水平设计下的置信区间宽度是否已宽到无法支撑组间推断。

    L1 网络状态:「该主题尚未进入 L1 对应网络」——CORRESPONDENCE_NETWORK.md 节点表把「相位 / 节律」列为物理侧节点(备注「周期性耦合的关键」)、「呼吸率 / 呼吸模式」列为生理侧观测项(备注「与节律耦合最直接」)、「气感报告」列为体验侧观测项,三者均未与易学侧任何节点连成边;节点表中完全没有「方差分量 / 变异源占比 / 信度系数 / 测量设计」这类测量学维度节点。边表 3.1–3.4 只登记五行↔五脏/五色/五音/五方/五季/五志、八卦↔五行、天干地支↔五行,无任何以变异源占比、组间方差结构或 G 理论为源或目标的边。EDGE_TABLE_science-yixue.md 仅有概念计数(丹田 51 篇 / 气功 48 篇 / 呼吸 189 篇),无层级(L1/L2/L3)与状态(active/open/conflict)判定。故本文封顶 C,所述机制均属方法学推演,不得读作已验证的生理机制。

    相关阅读

    证据等级: C
    参考文献: PMID 38806627, PMID 28882426, PMID 37143826, PMID 32666394, PMID 23412901, PMID 38695394, PMID 26843971, PMID 31460683, PMID 16474914, PMID 33818118, PMID 41549064, PMID 31517685, PMID 30898374, PMID 28628398, PMID 28641372, PMID 26808961


    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。

    \”Decreased Proportion of Body Position\” Does Not Equal \”More Stable Training\”: A Pitfall in Between-Group Comparisons Using Variance Component Structure

    Evidence level: C

    Decomposing total variance into “posture, belt, operator, individual, and residual” components, the most tempting step is comparing the five proportions between practitioners and novices. This step almost certainly leads to error because proportions are ratios: a decrease in the numerator or an increase in the denominator can both cause a drop.

    Main Text

    I. Proportions Are Ratios, Not Quantities

    The five variance sources must sum to 1. This constraint brings two consequences: any two proportions are naturally negatively correlated; a “posture proportion dropping from 40% to 25%” conveys no information on its own, because the posture variance could have dropped by 37%, or it could have remained unchanged while individual variance increased by 60%. The former indicates “reduced sensitivity to posture,” while the latter indicates “increased divergence between individuals.” These are opposite directions yet yield the same number.

    There is long-standing statistical consensus on component data: arithmetic means do not yield valid results within the simplex and require log-ratio transformation first (PMID 38806627); in share data, an absolute increase may manifest as a decrease in the share (PMID 28882426). Therefore, one should report absolute variance components, not proportions.

    II. Proportions Cannot Be Directly Used for Between-Group Inference

    The sampling distribution of proportions depends on the shared denominator across five components; none are independent. Running a t-test directly on proportions assumes the denominator is constant. In clinical trials, variance homogeneity tests have specialized methods—Jackknife, Cochran’s Q—which are more robust for heavy-tailed and skewed data (PMID 37143826); in mixed models, if variance varies by group, the correct approach is to model the variance directly (location-scale extensions), rather than only testing fixed effects (PMID 32666394).

    “Different variance structures between two groups” is a hypothesis about the variance function and must be tested at the variance level, not via descriptive comparisons of proportions.

    III. Posture Effects Do Not Exist in Everyone

    Healthy individuals may not have a posture effect to reduce at all. In 23 severe COPD patients, sitting showed thoracic asynchrony while supine eliminated it; upper thorax and abdomen were synchronous in sitting but strongly asynchronous in supine. In 12 age-matched healthy controls, neither posture showed any interval asynchrony (PMID 23412901).

    If both healthy practitioners and novices have posture variance near zero, a “lower posture proportion for practitioners” is an artifact of the denominator. This is a classic ceiling effect: the indicator has no variation space in the target population, making between-group comparisons meaningless. Before design, one must confirm that posture variance is significantly greater than zero in the population.

    IV. Small Designs Yield Unreliable Variance Component Estimates

    Posture has only 2 levels, and operators often have 1 to 2. Few measurements, short follow-ups, and large residual variances all expand the “non-regular” parameter region, making negative variances and boundary estimates the norm rather than anomalies (PMID 38695394); different estimators vary greatly in the probability of negative estimates (PMID 26843971).

    Inference is more complicated: asymptotic distributions for likelihood ratio, Wald, and Score tests fail at parameter boundaries, requiring permutation or Bayesian methods (PMID 31460683, PMID 16474914). Few raters themselves are a primary cause of ICC bias (PMID 33818118); confidence intervals for three modes may be so wide they span zero—”significant difference” and “no difference” become indistinguishable.

    V. What Kind of “Stability” Does Training Reduce?

    Skill research does support “greater stability for experts,” but one must clarify which kind. In 9 surgeons measuring 4 times each: junior residents had a score standard deviation of 5.8 (out of 30), while senior residents and attending physicians were both 1.8 (PMID 31517685); in simulator training, every 1% decrease in practice variance led to a 3.8-point increase on post-test and a 3.0-point increase in live transfer (PMID 30898374).

    However, these are repeated measures on the same individual, belonging to “residual + interaction” in crossed designs, not the “posture” design factor. A pistol shooting control group illustrates this better: expert COP displacement was smaller and multiscale entropy lower, yet both groups showed near-in-phase upper body-to-gun relative phase. The true difference lay in structure—experts had the gun lagging behind the body, novices had it leading (PMID 28628398). Stability does not mean every dimension shrinks; it means the structure shifts position.

    VI. Fundamental Trouble in Between-Group Comparison: No Random Assignment

    Practitioners and novices are not randomly assigned; they are self-selected. Those capable of training for ten years were likely already more able to sit still and control breathing—cross-sectional designs cannot distinguish “what is learned” from “who can persist.” Volunteer bias is systematic: UK Biobank participants are healthier and have better socioeconomic status than the general population (PMID 28641372).

    Another reading error must be guarded against. In SD-OCT measurements of children vs. adults: the two groups showed no difference in mean retinal thickness, but children’s COV was 3 to 4 times that of adults, and ICC was significantly lower (0.73–0.99 vs 0.986–0.993, PMID 26808961). Same mean, vastly different precision—this “between-group precision difference” is real yet most easily miswritten as “one group has greater ability.”

    Reversal: What Should Be Reported Are Not Proportions, But Two Absolute Quantities

    The error in the entire design lies in using a numerator proportion as a quality metric. What should truly be reported is: ① Intra-group reliability—calculate conditional reliability using G-theory’s D study for each group to see if the practitioner group truly has a higher D coefficient; ② Absolute residual variance—how much absolute noise actually decreased.

    There is also a known trap here: the classic G-theory assumes homoscedastic residuals, but actual data often violate this. Failing to model heterogeneity masks differences in measurement quality; one must calculate G and D conditionally (PMID 41549064).

    Therefore, if the practitioner group’s D coefficient did not rise but only the posture proportion dropped, it means variance merely shifted from “posture” to “person”—people trained for ten years differ more from each other. This is fully expected and unrelated to “more stable measurement.”

    Four Criteria: If posture absolute variance is significantly lower, D coefficient is higher, and absolute residual variance is also lower (all three aligned) ⇒ true robustness improvement; if posture absolute variance did not drop but proportion dropped ⇒ ratio artifact (denominator inflated), do not report “training reduces posture sensitivity”; if residual variance increased while D coefficient remained unchanged ⇒ variance merely moved locations, it is “a different kind of instability”; if both groups have posture absolute variance near zero ⇒ ceiling effect, this indicator is unusable for this population; switch to belt or operator factors.

    Three accompanying reports: report interactions Posture×Group, Belt×Group, Operator×Group, because between-group variance differences may simply mean a factor is ineffective for a specific group; when reporting variance components, always report confidence intervals and estimation methods; do not report p<0.05 for components falling on boundaries; calculate sample size based on “resolvable variance ratios,” not mean differences.

    Evidence and Limitations

    Verified: Closure constraints in component data lead to spurious correlations; arithmetic means of shares can exceed bounds (PMID 28882426); necessity of log-ratio transformation for share data (PMID 38806627); multiple methods for variance homogeneity tests and their performance differences in heavy-tailed/skewed data (PMID 37143826); need to model heterogeneity itself rather than only fixed effects in mixed models under heteroscedasticity (PMID 32666394); small measurement counts and short follow-ups expand negative variance/boundary regions (PMID 38695394); differences in negative estimation probabilities across estimators (PMID 26843971); classical tests fail at parameter boundaries requiring permutation or Bayesian methods (PMID 31460683, PMID 16474914); few raters cause ICC bias (PMID 33818118); G-theory requires extension to handle heteroscedasticity (PMID 41549064); differences in thoracic asynchrony between COPD and healthy controls regarding posture (PMID 23412901); decrease in individual variability with seniority among surgical residents (PMID 31517685); association between simulator practice variance reduction and skill transfer (PMID 30898374); coordination structure differences between expert and novice pistol shooters (PMID 28628398); volunteer bias in UK Biobank participants (PMID 28641372); same mean but different precision in OCT between children and adults (PMID 26808961).

    Partially Verified: The above skill variability evidence comes from surgical simulation training and shooting tasks; the “training” in Dantian breathing differs structurally in task design, feedback mechanisms, and assessment targets, serving only as directional reference; the between-group precision difference in OCT is an instrument measurement scenario, differing from “active human control of breathing”; evidence for between-group heterogeneity in posture effects comes from COPD and healthy controls, not covering long-term breathing training populations.

    To Be Verified: Whether there is a true difference in posture absolute variance between practitioners and novices in the Dantian breathing scenario (no direct research exists); whether practitioners truly have higher G/D coefficients; the magnitude of variance compression caused by “practitioners actively fixing posture”; whether Posture×Group interaction exists; whether confidence interval widths for five variance components under a 2-level design are already so wide as to preclude between-group inference.

    L1 Network Status: “This topic has not yet entered the L1 corresponding network”—the CORRESPONDENCE_NETWORK.md node table lists “Phase/Rhythm” as a physical-side node (note: “key for periodic coupling”), “Breathing Rate/Breathing Pattern” as a physiological-side observation (note: “most directly coupled to rhythm”), and “Qi Sensation Reports” as an experiential-side observation; none connect to any Yijing-side nodes. The node table contains no measurement dimension nodes such as “Variance Components/Variance Source Proportions/Reliability Coefficients/Measurement Design.” Edge tables 3.1–3.4 only register Five Phases↔Five Organs/Five Colors/Five Tones/Five Directions/Five Seasons/Five Emotions, Eight Trigrams↔Five Phases, Heavenly Stems/Earthly Branches↔Five Phases; no edges exist with variance source proportions, between-group variance structures, or G-theory as sources or targets. EDGE_TABLE_science-yixue.md only contains concept counts (Dantian 51 / Qigong 48 / Breathing 189), with no hierarchy (L1/L2/L3) or status (active/open/conflict) determination. Thus, this paper caps at Level C; described mechanisms belong to methodological deduction and should not be read as verified physiological mechanisms.

    Related Reading

    Evidence Level: C
    References: PMID 38806627, PMID 28882426, PMID 37143826, PMID 32666394, PMID 23412901, PMID 38695394, PMID 26843971, PMID 31460683, PMID 16474914, PMID 33818118, PMID 41549064, PMID 31517685, PMID 30898374, PMID 28628398, PMID 28641372, PMID 26808961


    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.

  • 「三个模态都说体位最大」——Kendall W 在这个设计里几乎证明不了什么

    Evidence level: C

    三个模态排序一致,只说明它们像;不说明它们对。而在这套设计里,连「像不像」都量不出来。

    正文

    一、先看这把尺子有几个刻度

    检验「体位是否主导变异」的常见做法,是让 RIP、FOT、EIT 三个模态各自给五个变异源(体位、带位、操作员、个体、残差)排序,再算 Kendall 一致性系数 W。k=3 个评判者,n=5 个对象。

    把这套设计的全部可能性穷举一遍——每个模态 5! = 120 种排序,三个模态合计 172.8 万种组合——能得到的 W 只有 44 个不同取值。α=0.05 的精确临界是 W ≥ 0.6889,44 个可达值里只有 13 个能判显著。

    n 更小时更糟:若只比三个源(体位、操作员、残差),可达 W 只剩 6 个值——0、0.111、0.333、0.444、0.778、1.0,其中仅 2 个能显著。

    含义很直接:W 不显著,不等于三个模态不一致。它可能只是这把尺子没有中间刻度。

    二、W 显著,也不等于排序站得住

    反过来的坑更深。W 只问「三个模态的排序像不像」,从不问「这个排序本身稳不稳」。

    做个模拟:设真实份额为体位 28%、带位 27%(几乎并列),即使有效重复次数涨到 500,体位被误判成非第一的概率仍有 36%。若真实差拉大到 40% 对 20%,重复 30 次就能把误判率压到 7.4%。

    W = 1.0 完全可以是三个模态一起在噪声里抖。 一致性(agreement)和正确性(accuracy)是两件事,前者可以完美,后者可以全错。

    三、份额是成分数据,五个秩不独立

    变异源占比必须加起来等于 100%,这叫闭合数据(closed data)。闭合约束会让各份额天然负相关:一个涨,别的必须跌。

    有研究用垃圾组分数据演示过后果:同一批数据按绝对质量算,可避免蔬菜垃圾与塑料包装正相关;改用质量分数算,相关变成负的(PMID 28882426);算术均值甚至能给出负的比例,置信区间 (-4.03, 8.45)。

    这直接打掉 W 的前提——W 的零分布假设 k 个排序彼此独立,而成分数据的五个秩被「总和恒定」绑在一起。正确做法是先做对数比变换(clr/ilr)再谈方差(PMID 38806627)。

    四、水平数太少,方差分量自己在飘

    体位只有 2 个水平(坐/卧),带位 2-3 个,操作员 2-3 个。水平少是所有麻烦的源头。

    负的方差估计和边界估计在这种设计里很常见。有方法学工作专门讨论「伪方差」:某些数据的边际分布与「负方差随机截距斜率模型」在数学上完全等同,而测量次数少、随访短会扩大这片区域(PMID 38695394)。另有模拟比较各类估计量「出现负估计的概率」(PMID 26843971)。

    检验「某方差分量为零」更麻烦:零落在参数空间边界上,似然比与得分统计量的常规卡方分布是错的,得靠置换法近似有限样本分布(PMID 22930674)。

    而 REML 的抽样方差走大样本渐近——有研究要样本量到 1413 才认为渐近近似够用(PMID 35197554)。二十几个人的样本里,份额的抽样误差很可能大于要比较的那点差距。

    五、「有体位效应」≠「体位是最大源」

    体位影响呼吸的证据其实很扎实:系统综述 43 项研究,立位到卧位功能残气量显著下降(PMID 30305051);62 人 IOS 测试,卧位 R5、R5-R20 全升、X5 更负,肥胖组更明显(PMID 36787314);17 例 ARDS 患者坐起 30 分钟后,EIT 的 ROI 通气分布整体重排(PMID 39116511)。

    但这些都只说明体位有效应,不说明体位是最大的源。中间隔着一个跨源头比大小,而这恰恰是最不稳的一步。

    顺带一提:EIT 区域通气分布在固定通气设置下重复性极好(10 头猪,通气重心随 PEEP 从 45±3% 移到 49±3%,区间极窄,PMID 17664640)。重复性好只代表同一体位内稳,与跨源头比大小无关。

    六、还有一个隐藏前提:异方差

    G 理论经典假设各条件下残差方差齐性。生理数据里这个假设常不成立。有工作用位置-尺度混合模型扩展 G 理论,发现不建模异方差会掩盖真实的测量质量差异,须按条件分别计算 G 与 D 系数才看得出来(PMID 41549064)。多层级评分者信度同理——评判者数量少是 ICC 偏倚与低效率的主因(PMID 33818118)。

    三个模态就是三个评判者,本来就少;再叠上异方差,「平均份额」这个数代表不了任何单一条件。

    反转:该报的不是 W,是相邻差值的区间

    整个设计的毛病是问错了问题。真正要答的是「体位这份额比第二名大多少,这个差有没有超出不确定度」,而不是「三个模态的排序像不像」。做法是把排序换成两两差值加 bootstrap 置信区间:只有 CI 不跨零的相邻对,才算分得出高低。W 一个这样的问题都答不了。

    四条判法:三模态 W 均显著且相邻差 CI 全不跨零 ⇒ 排序可信,可谈体位主导;W 显著但相邻差 CI 跨零 ⇒ 只能报「体位高于其余」,不能报完整排序;W 不显著但某两源差 CI 稳定不跨零 ⇒ 是粒度问题不是不一致,改报差值;W 不显著且所有差 CI 都跨零 ⇒ 该数据支撑不了任何排序结论,「体位主导」退回假设。

    配套三条:报 W 必须同时报可达值个数与临界值在其中的排位;成分数据先 ilr/clr 变换;样本量按「要分辨的最小份额差」反解,别按「够不够做 W」。

    证据与局限

    已验证: Kendall W 在 k=3、n=5 时的可达取值离散性与精确排列临界值(本文穷举计算);闭合数据(成分数据)会引入伪相关、算术均值可给出负比例(PMID 28882426);对数比变换处理成分数据的必要性(PMID 38806627);负方差与伪方差模型的成因,小测量次数与短随访扩大该区域(PMID 38695394);零方差检验在参数边界上常规卡方分布失效、需置换法(PMID 22930674);REML 方差分量抽样方差的渐近性对小样本可能有偏(PMID 35197554);体位对肺功能的影响(系统综述 43 项,PMID 30305051);IOS 卧位上阻抗升高、X5 更负(PMID 36787314);EIT 坐位改变通气分布(PMID 39116511);EIT 区域通气分布重复性(PMID 17664640);异方差下 G 理论需扩展(PMID 41549064);评判者数量少导致 ICC 偏倚(PMID 33818118)。

    部分验证: 上述体位证据多来自机械通气患者、肥胖者或 ARDS 人群,健康修炼者在半卧/端坐下的体位效应幅度是否同量级未知;份额排序的翻转概率为本文蒙特卡洛模拟结果,依赖设定的真实份额与误差结构,非实测;成分数据在变异源拆分场景下的伪相关幅度未直接测量。

    待验证: 丹田呼吸场景下五个变异源的真实份额范围(无任何直接研究);体位与带位是否存在交互(卧位腹带受压后张力变化是否简单可加);三模态各自的残差方差是否满足齐性;把 W 换成差值区间后,原「体位主导」结论是否仍成立。

    L1 网络状态:「该主题尚未进入 L1 对应网络」——CORRESPONDENCE_NETWORK.md 节点表把「相位 / 节律」列为物理侧节点、「呼吸率 / 呼吸模式」列为生理侧观测项,均未与易学侧任何节点连成边;节点表中完全没有「变异源 / 方差分量 / 一致性系数 / 排序」这类测量学维度节点,边表 3.1–3.4 只登记五行↔五脏/五色/五音/五方/五季/五志、八卦↔五行、天干地支↔五行,无任何以变异源拆分、体位主导或 Kendall W 为源或目标的边。EDGE_TABLE_science-yixue.md 仅有概念计数(丹田 51 篇 / 气功 48 篇 / 呼吸 189 篇),无层级与状态判定。故本文封顶 C。

    相关阅读

    证据等级: C
    参考文献: PMID 28882426, PMID 38806627, PMID 38695394, PMID 26843971, PMID 22930674, PMID 35197554, PMID 30305051, PMID 36787314, PMID 39116511, PMID 17664640, PMID 41549064, PMID 33818118


    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。

    \”Three Modalities Say Posture is Biggest\” — Kendall W Can’t Prove Much in This Design

    Evidence level: C

    All three modalities rank consistently; this only indicates similarity, not correctness. In this design, even the degree of “similarity” is unmeasurable.

    Main Text

    I. First, How Many Tick Marks Does This Ruler Have?

    A common approach to testing whether “posture dominates variation” is to have RIP, FOT, and EIT each rank five sources of variation (posture, belt position, operator, individual, residual), then calculate Kendall’s coefficient of concordance W. With $k=3$ raters and $n=5$ objects.

    Exhaustively enumerate all possibilities for this design—each modal has $5! = 120$ rankings, totaling over 1.7 million combinations for three modalities—only 44 distinct values of W can be obtained. The exact critical value at $\alpha=0.05$ is $W \ge 0.6889$. Of the 44 attainable values, only 13 are significant.

    It is worse when $n$ is smaller: if comparing only three sources (posture, operator, residual), the attainable W values drop to just 6: 0, 0.111, 0.333, 0.444, 0.778, and 1.0. Only 2 of these are significant.

    The implication is direct: Non-significant W does not mean the three modalities are inconsistent. It may simply mean this ruler lacks intermediate tick marks.

    II. Significant W Does Not Mean the Ranking Is Stable

    The reverse pitfall is deeper. W only asks “how similar are the three rankings,” never “is this ranking itself stable.”

    Run a simulation: Assume true proportions are 28% for posture and 27% for belt position (almost tied). Even with effective replication increasing to 500, the probability of misjudging posture as non-first remains 36%. If the true difference widens to 40% vs. 20%, only 30 replications are needed to reduce the misjudgment rate to 7.4%.

    W = 1.0 can simply mean all three modalities are shaking in the noise. Agreement and accuracy are two separate matters; the former can be perfect while the latter is entirely wrong.

    III. Proportions Are Compositional Data; Five Ranks Are Not Independent

    The proportions of variation sources must sum to 100%; this is closed data. The closure constraint naturally induces negative correlation among shares: as one increases, others must decrease.

    Studies have demonstrated consequences using compositional data on garbage components: calculating by absolute mass avoids a positive correlation between vegetable waste and plastic packaging; switching to mass fractions turns the correlation negative (PMID 28882426); even the arithmetic mean can yield negative proportions with a confidence interval of (-4.03, 8.45).

    This directly invalidates W’s premise—the assumption that $k$ rankings are independent. The five ranks of compositional data are bound together by the “constant sum” constraint. The correct approach is to perform log-ratio transformations (clr/ilr) before discussing variance (PMID 38806627).

    IV. Too Few Levels Cause Variance Components to Float

    Posture has only 2 levels (sitting/lying), belt position 2-3, and operator 2-3. Few levels are the source of all troubles.

    Negative variance estimates and boundary estimates are common in such designs. Methodological work specifically discusses “pseudo-variance”: certain data marginal distributions are mathematically equivalent to a “negative variance random intercept slope model,” and few measurements plus short follow-ups expand this region (PMID 38695394). Other simulations compare estimators regarding “the probability of yielding negative estimates” (PMID 26843971).

    Testing “whether a variance component is zero” is more complicated: zero lies on the boundary of the parameter space, so the standard chi-square distribution for likelihood ratio and score statistics is incorrect; permutation methods are needed to approximate the finite sample distribution (PMID 22930674).

    REML sampling variance relies on large-sample asymptotics—studies suggest sample sizes up to 1413 are needed for the asymptotic approximation to be adequate (PMID 35197554). In a sample of only dozens, the sampling error for proportions may exceed the small differences being compared.

    V. “Posture Effect Exists” $\neq$ “Posture Is the Largest Source”

    Evidence that posture affects breathing is actually quite solid: a systematic review of 43 studies found significant decreases in functional residual capacity from sitting to lying (PMID 30305051); in a 62-person IOS test, lying showed increased R5 and R5-R20 with more negative X5, especially in the obese group (PMID 36787314); in 17 ARDS patients, sitting for 30 minutes caused a global rearrangement of EIT ROI ventilation distribution (PMID 39116511).

    However, these only show posture has an effect, not that it is the largest source. Separating them requires cross-source comparison, which is precisely the most unstable step.

    Incidentally: EIT regional ventilation distribution shows excellent repeatability under fixed ventilation settings (10 pigs; ventilation centroid shifted from $45 \pm 3\%$ to $49 \pm 3\%$ with a very narrow interval, PMID 17664640). Good repeatability only indicates stability within the same posture, unrelated to cross-source magnitude comparison.

    VI. Another Hidden Premise: Heterogeneity of Variance

    The G-theory classic assumption is homoscedasticity of residuals across conditions. This assumption often fails in physiological data. Work using location-scale mixed models extended G-theory, finding that failing to model heterogeneity masks true differences in measurement quality; G and D coefficients must be calculated separately by condition to reveal them (PMID 41549064). Multi-level rater reliability is similar—few raters are the main cause of ICC bias and low efficiency (PMID 33818118).

    Three modalities are three raters, already few; stacking heterogeneity means the “average proportion” represents nothing for any single condition.

    Reversal: What Should Be Reported Is Not W, But Confidence Intervals for Adjacent Differences

    The flaw in the entire design is asking the wrong question. The real question should be “by how much is the posture proportion larger than the second place, and does this difference exceed uncertainty,” not “how similar are the three rankings.” The solution is to replace rankings with pairwise differences plus bootstrap confidence intervals: only adjacent pairs with CIs not crossing zero count as distinguishable. W cannot answer such a question.

    Four judgment criteria:
    1. All three-modal W significant AND all adjacent difference CIs non-crossing zero $\Rightarrow$ ranking is credible; posture dominance can be discussed.
    2. W significant BUT adjacent difference CIs cross zero $\Rightarrow$ only report “posture higher than others,” not a full ranking.
    3. W non-significant BUT some source pair difference CI stable non-crossing zero $\Rightarrow$ it is a granularity issue, not inconsistency; report differences.
    4. W non-significant AND all difference CIs cross zero $\Rightarrow$ data does not support any ranking conclusion; “posture dominance” reverts to hypothesis.

    Three supporting rules: Always report the number of attainable values and their rank relative to the critical value when reporting W; perform ilr/clr transformation for compositional data first; calculate sample size based on the “minimum proportion difference to resolve,” not based on “enough for W.”

    Evidence and Limitations

    Verified: Kendall W attainable value discreteness and exact permutation critical values for $k=3, n=5$ (exhaustive calculation in this paper); compositional data introduces pseudo-correlation and arithmetic means can yield negative proportions (PMID 28882426); necessity of log-ratio transformations for compositional data (PMID 38806627); causes of negative and pseudo-variance models, where few measurements and short follow-ups expand this region (PMID 38695394); standard chi-square distribution failure for zero-variance tests at parameter boundaries requiring permutation (PMID 22930674); REML sampling variance asymptotics may be biased for small samples (PMID 35197554); posture effects on lung function (systematic review of 43 studies, PMID 30305051); IOS impedance increases and X5 becomes more negative in lying position (PMID 36787314); EIT alters ventilation distribution in sitting position (PMID 39116511); EIT regional ventilation distribution repeatability (PMID 17664640); G-theory requires extension under heterogeneity (PMID 41549064); few raters cause ICC bias (PMID 33818118).

    Partially Verified: Most posture evidence comes from mechanically ventilated patients, obese individuals, or ARDS populations; the magnitude of posture effects in semi-recumbent/sitting healthy practitioners remains unknown; flipping probabilities for proportion rankings are Monte Carlo simulation results from this paper, dependent on assumed true proportions and error structures, not measured; magnitude of pseudo-correlation in compositional data for variation source decomposition was not directly measured.

    To Be Verified: True proportion ranges for the five variation sources in the Dantian breathing scenario (no direct research exists); whether posture and belt position interact (whether tension changes after lying with abdominal belts are simply additive); whether residual variances for the three modalities satisfy homogeneity; whether replacing W with difference intervals preserves the original “posture dominance” conclusion.

    L1 Network Status: “This topic has not yet entered the L1 corresponding network”—The CORRESPONDENCE_NETWORK.md node table lists “Phase/Rhythm” as physical-side nodes and “Respiration Rate/Pattern” as physiological-side observations, but neither connects to any Yijing nodes; the node table contains no measurement-dimension nodes such as “Variation Source/Variance Component/Concordance Coefficient/Ranking,” and Edge Tables 3.1–3.4 only register Five Phases $\leftrightarrow$ Five Organs/Five Colors/Five Tones/Five Directions/Five Seasons/Five Emotions, Eight Trigrams $\leftrightarrow$ Five Phases, Heavenly Stems/Branches $\leftrightarrow$ Five Phases. No edges exist with variation source decomposition, posture dominance, or Kendall W as source or target. EDGE_TABLE_science-yixue.md only contains concept counts (Dantian 51 / Qigong 48 / Breathing 189), with no hierarchy or status determination. Thus, this paper is capped at Level C.

    Related Reading

    Evidence Level: C
    References: PMID 28882426, PMID 38806627, PMID 38695394, PMID 26843971, PMID 22930674, PMID 35197554, PMID 30305051, PMID 36787314, PMID 39116511, PMID 17664640, PMID 41549064, PMID 33818118


    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.

  • MDD 不是「最小有用差异」,是「最小非噪声差异」:丹田 CoV 效应量的信噪比门槛该怎么定

    Evidence level: C

    一个参数值不值得当终点,不看它「能不能变」,而看它「变了之后能不能和抖动分开」——而这条线,比大多数人以为的要高。

    正文

    一、MDD 回答的问题,和它不回答的问题

    MDD(最小可检测差异)是从 SEM(测量标准误)推出来的:MDC95 ≈ 1.96 × √2 × SEM ≈ 2.77 × SEM。它只回答一件事——这个数变了,是不是仪器自己在抖。它完全不管这个变化有没有生理意义。

    问题在于,MDD 这个数本身的质量没人担保。2025 年一项系统综述纳入 49 项功能测试的信度研究,报告了 ICC、SEM、MDC,但用 COSMIN 清单评估后全部被评为「不合格」(PMID 40895367)。也就是说,文献里那些看起来很精确的 MDD 数字,方法学底子普遍是虚的。

    二、CoV 尤其不适合直接套 MDD

    CoV = SD / 均值,是个比值。它的抽样分布依赖均值大小:均值越接近零,CoV 越爆炸;同一个绝对变异,在低均值个体身上会被算得更大。这意味着「CoV 的 MDD」不是一个跨人群通用的常数——它随基线水平漂移。

    更麻烦的是组间比较。CoV 不能像均值那样直接做 t 检验,必须走对数变换(PMID 10588295)。所以「修炼者 CoV 的 MDD」和「新手 CoV 的 MDD」本身就是两个不同的数,拿同一个 MDD 去卡两组,等于用一把会伸缩的尺子量两样东西。

    三、真正成熟的框架是 RCV,不是 MDD

    临床检验早有更完整的解法:参考变化值 RCV = √2 · Z · √(CV_A² + CV_I²),把分析变异(CV_A)和个体内生物学变异(CV_I)分开算(PMID 21958344)。关键在于 CV_I 必须单独估,而且它对不同指标差一个数量级。

    实测数据很直白:同一批慢性肾病患者、同一时间窗内,肾小球滤过标记物的个体内 CV 只有 5.4%–7.4%,而白蛋白尿的 CV 高达 29.7%–50.6%(PMID 30031564)。同样的人、同样的重复测量设计,不同指标的「最小可辨变化」差十倍。这直接否掉了「给所有参数定一个统一 MDD」的做法——SnR 门槛必须逐参数定。

    四、SnR > 2 这个数字,出处是空的

    「效应量要大于 MDD 的 2 倍」听起来像行业惯例,但它没有生理学依据。一项针对 380 位 RCT 通讯作者的调查显示:选「要检测多大差值」根本没有指南,中位选择是疼痛控制改善 7.0%、死亡率或复发降低 5.0%,而且随基线风险显著变化(PMID 27473336)。连「要检测多大」都是主观拍定的,「2 倍」自然只是拍定值的拍定值。

    更糟的是预期效应量系统性偏高。194 项 RCT 的分析发现:阴性结果组预期效应量比实测高 42%(95%CI 32–51),post hoc 检验力只有 20.2%(PMID 26041717)。也就是说,用预实验效应量去除 MDD 得到的 SnR,分子天生偏大。

    五、个体反应异质性会让平均效应量失真

    就算 SnR 算出来达标,也要小心它掩盖了什么。对最大摄氧量训练反应的批判性综述指出:个体间反应差异极大,存在明确的「高反应者」与「无反应者」(PMID 28097487)。用一个平均效应量去除以 MDD,得到的 SnR 描述的是一个不存在的「平均人」。

    呼吸侧的信度数据也提醒同一件事:可穿戴设备测呼吸率,静息 ICC 0.76–0.84(PMID 30694966),看起来很好;但膈肌功能的多指标评估里,峰值类指标 ICC 0.70–0.94,斜率类指标 ICC ≤ 0.20(PMID 41706258)。同一台设备、同一批人,指标形态就决定了信度。若 CoV 的算法里带拟合斜率,直接报废。

    六、技术噪声地板有个参照量级

    34 个实验室、19 个国家测同一批标准物质:实验室内 CV ≤ 4.2%,实验室间 CV < 14%(PMID 39362843)。这是「纯技术噪声」的量级参照。另外,MRI 换机后重测信度与样本量估计都必须重算(PMID 32032737)——换中心就要重算 MDD,没有继承这回事。

    反转:SnR 卡的不是参数,是设计

    如果某参数效应量只有 MDD 的 1.5 倍,正确反应不是「换个参数」,而是增加重复测量次数——SEM 按 √n 下降,MDD 跟着降。把 SnR > 2 写成不带重复次数的固定门槛,等于用淘汰参数来掩盖设计不足。四个格子的判法:全部候选 SnR < 2 ⇒ 这个试验不该做;仅 CoV 达标而 %AB 不达标 ⇒ 疑 CoV 的 MDD 被低估;全部达标但个体异质性大 ⇒ 终点改「响应者比例」;达标但跨中心重算后掉到 2 以下 ⇒ 中心效应吃掉效应,先统一协议。

    证据与局限

    已验证: RCV 框架的数学形式与 CV_A/CV_I 分离原则(PMID 21958344);个体内 CV 的指标间数量级差异(PMID 30031564);MDC 与 SEM 的换算关系及功能测试 MDC 的广泛报告(PMID 18356292, PMID 40895367);呼吸率与膈肌功能指标的 ICC 分层(PMID 30694966, PMID 41706258);实验室间 CV 基准(PMID 39362843);换设备后信度与样本量需重估(PMID 32032737)。

    部分验证: SnR > 2 作为终点入选门槛在方法学上合理但无统一出处(PMID 27473336 显示「要检测多大」本身无指南);预期效应量系统性高估已有明确证据(PMID 26041717),但高估幅度是否适用于练功类干预未知;CoV 的组间比较须对数变换(PMID 10588295)在呼吸参数上未直接验证。

    待验证: 丹田呼吸场景下 CoV 的个体内 CV 究竟是多少(无任何直接研究);修炼者与新手 CoV 的 MDD 是否真的不同;「气沉」效应量在不同重复次数下的 SnR 曲线;响应者比例终点是否比均值终点更敏感。

    相关阅读

    证据等级: C
    参考文献: PMID 40895367, PMID 21958344, PMID 30031564, PMID 26041717, PMID 28097487, PMID 10588295, PMID 18356292, PMID 30694966, PMID 41706258, PMID 39362843, PMID 32032737, PMID 27473336

    L1 网络状态:「该主题尚未进入 L1 对应网络」——CORRESPONDENCE_NETWORK.md 节点表只把「相位 / 节律」列为物理侧节点(备注「周期性耦合的关键」)、把「呼吸率 / 呼吸模式」列为生理侧观测项(备注「与节律耦合最直接」),二者均未与易学侧任何节点连成边;节点表中完全没有「变异系数 / 最小可检测差异 / 信噪比 / 测量误差」这类测量学维度节点,边表 3.1–3.4 只登记五行↔五脏/五色/五音/五方/五季/五志、八卦↔五行、天干地支↔五行,无任何以丹田参数、CoV、MDD 或终点选择为源或目标的边。EDGE_TABLE_science-yixue.md 仅有概念计数(丹田 51 篇 / 气功 48 篇 / 呼吸 189 篇),无层级(L1/L2/L3)与状态(active/open/conflict)判定。故本文封顶 C。


    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。

    MDD Is Not \”Minimum Useful Difference\” but \”Minimum Non-Noise Difference\”: How to Set the Signal-to-Noise Ratio Threshold for Dantian CoV Effect Sizes

    Evidence level: C

    A parameter value is not worth serving as an endpoint based on whether it “can change,” but rather on whether it “can be distinguished from noise after changing”—and this threshold is higher than most assume.

    Main Text

    I. The Question MDD Answers and the Questions It Does Not Answer

    MDD (Minimum Detectable Difference) is derived from SEM (Standard Error of Measurement): MDC95 ≈ 1.96 × √2 × SEM ≈ 2.77 × SEM. It addresses only one thing: if this number changes, is the instrument itself fluctuating? It completely ignores whether the change has physiological significance.

    The issue lies in the quality of the MDD number itself, which is not may support. A 2025 systematic review including 49 reliability studies from functional tests reported ICC, SEM, and MDC, but after assessment using the COSMIN checklist, all were rated “inadequate” (PMID 40895367). In other words, the seemingly precise MDD numbers in the literature generally rest on a weak methodological foundation.

    II. CoV Is Especially Unsuitable for Direct Application of MDD

    CoV = SD / Mean is a ratio. Its sampling distribution depends on the mean magnitude: as the mean approaches zero, CoV explodes; the same absolute variation is calculated as larger in individuals with lower baseline means. This means the “CoV MDD” is not a universal constant across populations—it drifts with baseline levels.

    Complications arise in between-group comparisons. CoV cannot undergo a t-test like the mean; it requires a logarithmic transformation (PMID 10588295). Therefore, the “MDD for CoV in practitioners” and the “MDD for CoV in novices” are fundamentally different numbers. Applying a single MDD to compare two groups is equivalent to using a stretching ruler to measure two different things.

    III. The Mature Framework Is RCV, Not MDD

    Clinical diagnostics already have a more complete solution: the Reference Change Value (RCV) = √2 · Z · √(CV_A² + CV_I²), separating analytical variation (CV_A) from individual biological variation (CV_I) (PMID 21958344). The key is that CV_I must be estimated separately, and it differs by an order of magnitude across different indicators.

    The empirical data is straightforward: in the same cohort of chronic kidney disease patients within the same time window, the individual CV for glomerular filtration markers was only 5.4%–7.4%, whereas the CV for albuminuria reached 29.7%–50.6% (PMID 30031564). The same people, with the same repeated measurement design, show a “minimum discernible change” differing tenfold across different indicators. This directly invalidates the practice of setting a unified MDD for all parameters—the Signal-to-Noise Ratio (SnR) threshold must be determined per parameter.

    IV. The SnR > 2 Figure Has No Source

    The statement “effect size must be greater than twice the MDD” sounds like industry convention, but it lacks physiological basis. A survey of 380 RCT correspondence authors found: there is no guideline on “how large a difference to detect,” with the median selection being a 7.0% improvement in pain control or a 5.0% reduction in mortality or recurrence, varying significantly with baseline risk (PMID 27473336). Even “how large to detect” is subjectively set; thus, “twice” is merely the square of a subjective setting.

    Worse still, expected effect sizes are systematically overestimated. Analysis of 194 RCTs found that expected effect sizes in negative-result groups were 42% higher than observed (95%CI 32–51), with post hoc power only at 20.2% (PMID 26041717). In other words, the SnR derived by dividing the pre-experimental effect size by MDD has a molecularly inflated numerator.

    V. Individual Response Heterogeneity Distorts Average Effect Sizes

    Even if SnR calculates as meeting the standard, one must be careful what it masks. A critical review of responses to VO2 max training indicates extreme inter-individual variability, with clear “high responders” and “non-responders” (PMID 28097487). Using an average effect size divided by MDD describes a non-existent “average person.”

    Reliability data from respiratory side measurements also warns of this: wearable devices measuring respiratory rate show resting ICC 0.76–0.84 (PMID 30694966), which appears good; however, in multi-indicator assessments of diaphragm function, peak-type indicators show ICC 0.70–0.94, while slope-type indicators show ICC ≤ 0.20 (PMID 41706258). The same device and same cohort yield different reliability based on indicator morphology. If the CoV algorithm includes fitted slopes, it should be discarded immediately.

    VI. Technical Noise Floors Have a Reference Magnitude

    34 laboratories across 19 countries measured the same batch of standard substances: intra-laboratory CV ≤ 4.2%, inter-laboratory CV < 14% (PMID 39362843). This is the reference magnitude for “pure technical noise.” Additionally, re-test reliability and sample size estimation after MRI scanner changes must be recalculated (PMID 32032737)—changing centers requires recalculating MDD; there is no inheritance.

    Reversal: SnR Filters Design, Not Parameters

    If a parameter’s effect size is only 1.5 times the MDD, the correct response is not “change the parameter,” but increase the number of repeated measurements—SEM decreases by √n, and MDD follows. Writing SnR > 2 as a fixed threshold without repeated counts equals using the elimination of parameters to mask design deficiencies. The four-quadrant judgment method: all candidate SnR < 2 ⇒ the trial should not proceed; only CoV meets standard while %AB fails ⇒ suspect CoV MDD is underestimated; all meet standards but high individual heterogeneity ⇒ change endpoint to “proportion of responders”; meets standards but drops below 2 after cross-center recalculation ⇒ center effects consume the effect, first unify protocols.

    Evidence and Limitations

    Verified: Mathematical form of the RCV framework and the separation principle of CV_A/CV_I (PMID 21958344); order-of-magnitude differences in individual CV across indicators (PMID 30031564); conversion relationship between MDC and SEM and widespread reporting of functional test MDC (PMID 18356292, PMID 40895367); stratification of ICC for respiratory rate and diaphragm function indicators (PMID 30694966, PMID 41706258); inter-laboratory CV benchmarks (PMID 39362843); need for re-estimation of reliability and sample size after equipment changes (PMID 32032737).

    Partially Verified: SnR > 2 as an endpoint inclusion threshold is methodologically reasonable but lacks a unified source (PMID 27473336 shows no guideline on “how large to detect”); systematic overestimation of expected effect sizes has clear evidence (PMID 26041717), but whether the overestimation magnitude applies to training interventions is unknown; between-group comparison of CoV requires logarithmic transformation (PMID 10588295), not directly verified for respiratory parameters.

    To Be Verified: What the individual CV for CoV is specifically in the Dantian breathing scenario (no direct research exists); whether MDD for CoV truly differs between practitioners and novices; the SnR curve of “Qi Sinking” effect size across different repetition counts; whether a proportion-of-responder endpoint is more sensitive than a mean endpoint.

    Related Reading

    Evidence Level: C
    References: PMID 40895367, PMID 21958344, PMID 30031564, PMID 26041717, PMID 28097487, PMID 10588295, PMID 18356292, PMID 30694966, PMID 41706258, PMID 39362843, PMID 32032737, PMID 27473336

    L1 Network Status: “This topic has not entered the L1 corresponding network”—the CORRESPONDENCE_NETWORK.md node table lists only “Phase/Rhythm” as a physical-side node (note: “key for periodic coupling”) and “Respiratory Rate/Pattern” as a physiological-side observation item (note: “most directly coupled with rhythm”), but neither connects to any node on the Yijing side; there are no nodes in the node table with measurement dimensions such as “Coefficient of Variation/Minimum Detectable Difference/Signal-to-Noise Ratio/Measurement Error,” and Edge Tables 3.1–3.4 only register Five Phases ↔ Five Organs/Five Colors/Five Tones/Five Directions/Five Seasons/Five Emotions, Eight Trigrams ↔ Five Phases, Heavenly Stems and Earthly Branches ↔ Five Phases, with no edges sourced or targeted by Dantian parameters, CoV, MDD, or endpoint selection. EDGE_TABLE_science-yixue.md only contains concept counts (Dantian 51 articles / Qigong 48 articles / Breathing 189 articles), with no hierarchy (L1/L2/L3) or status (active/open/conflict) determination. Therefore, this article is capped at Level C.


    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.

  • 「三模态都朝一个方向变」这件事,其实没人验过:丹田呼吸跨中心方向一致性检验

    Evidence level: C

    三角验证的价值不在于三个模态都点头,而在于每一个都可以独立地摇头。

    正文

    上一篇给 %AB、φ、X5、CoV 定了”通用尺子”——ICC 多少、MDD 多大。但尺子准,不等于量出来的方向对。这是两套完全独立的证伪。 一台仪器可以稳定地高估 10%,ICC 照样到 0.9;它量什么都是同一个偏,前后一致,却系统性偏离真值。

    这一点在丹田参数上尤其危险,因为效应量本来就小。EIT 测潮气量的最新人体研究给了个精确的警示:27 例受试者个体内阻抗-容积回归 R² 平均 0.89,看着漂亮;但把人群合起来,R² 掉到 0.485(PMID 41091340)。个体能对上、群体对不上——跨中心合并数据时踩的正是这一脚。更麻烦的是偏倚常常不是常数,而是随数值增大而放大(比例偏倚),在”变化量只有几个百分点”的参数上,它会伪装成”响应”。

    第二条:三个模态同向,这个假设本身在成熟文献里就已经被打破过一次。早产儿与足月儿的经典对照(PMID 9387956):相位角 60.6° vs 12.5°(p<0.0001)、呼吸做功指数 1.35 vs 1.01(p=0.001),两项都分得极开;可同一批数据里的胸廓贡献比 25.5% vs 36.3%,p=0.11,毫无差异。同一构造、同一批人、同一个方向该变的地方——%AB 就是不动。所以”三模态应当同向”不是一个可以默认的前提,而是一个需要逐个检验的假设。

    第三条,跨模态一致的合格线,必须按”跨方法”来估,不能按”同方法重复测”来估。测量不变性的分级讲得很清楚:结构一致(configural)、载荷一致(metric)之后,还要过截距一致(scalar),才能比较均值(PMID 26387118)。很多指标卡在前两级:相关很强、排序也稳,但截距不同,均值压根不在同一把尺上。这正是丹田三模态最可能的处境——都朝”更好”的方向动,但”好了多少”不可比。

    第四条,中心效应是有结构的噪声,不是随机的。237 个中心、9978 例的随机对照数据显示:中心间结局的 OR 95% 范围宽达 0.27–3.71(相差十三倍),可治疗效应几乎没动(1.22→1.24,PMID 21867540)。这个结果两面都扎人:好的一面是”中心间基线差异大”不等于”效应方向会翻”;坏的一面是它诱人得出”中心无关紧要”,从而根本不去做方向检验。该检的是”效应×中心”交互,不是”基线×中心”差异。

    第五条,方向翻转在方法学里有正式名字:替代终点悖论——标志物与结局正相关,处理却对两者产生反向效应(PMID 25236906)。真实代价有据可查:抗心律失常药有效减少室性早搏,却显著增加死亡(PMID 11714414)。更隐蔽的是,同一批数据换成”绝对变化”或”响应者比例”两种终点,结论就能打架,且极端的切点值最容易制造冲突(PMID 37274016)。

    第六条,一致性是在安静条件下成立的。RIP 与光电体积描记测胸腹异步,静息和运动下都落在 Bland-Altman 一致界限内(PMID 28767680)——这是正面证据。但换成 10 名成人做自然呼吸、运动伪迹、模拟气道阻塞、打鼾、咳嗽:没有任何一个传感器在所有动作下都最准,自然呼吸时传感器间互相关很高,一进伪迹段就掉(PMID 9210815)。RIP 的校准更脆:换个体位后,等容 maneuver 校准只有 38%–65% 的测值还落在 ±10% 内(PMID 7091869);而免校准的 QDC 法在数学校验下只在潮气量恒定或近似恒定时才给对系数,自发呼吸下结果严重依赖选段标准(PMID 11181615)。练功态恰恰是”条件变了”的那一侧:呼吸更慢、幅度更大、腹壁主动用力。

    最后,三模态的”方向”本来不是同一个方向。%AB 是相对量,分母一动它就动;φ 是跨零角度;CoV 是空间分布;而 X5 强依赖肺容积,慢深呼吸本身就把容积轨迹改了(PMID 20335298)。X5 跟着变,是容积的连带效应,不是”气沉”的独立证据。

    所以这个检验的样本单位是中心×模态的格子,不是受试者。3 个中心 × 4 个模态只有 12 格,一个方向翻转就足以把”全同向”打成”7/12″。先算清这个格子数够不够做方向投票,再谈收集数据——这比事后解释”为什么某个中心不一样”省得多。

    证据与局限

    已验证: 同一构造下不同参数可给出相反结论(PMID 9387956);个体内高相关不等于群体可用(PMID 41091340);中心间基线异质不必然改变效应方向(PMID 21867540);替代终点可方向翻转且造成实际伤害(PMID 25236906、PMID 11714414);同终点不同口径可致结论冲突(PMID 37274016);RIP 校准对体位与呼吸条件高度敏感(PMID 7091869、PMID 11181615);传感器一致性在伪迹条件下降级(PMID 9210815)。

    部分验证: RIP 与 OEP 在静息与运动下一致性良好(PMID 28767680),说明”两法可比”在受控条件下可以成立;测量不变性框架给出了”何时才能比均值”的判定阶梯(PMID 26387118)。

    待验证: ① 同一批修炼者跨 ≥3 中心轮转后,%AB/φ/X5/CoV 的效应方向是否同号?② 方向一致率是否显著高于随机(方向投票检验)?③ 方向翻转的模态能否被协议变量(体位、校准法、带位、采集时长)解释?④ 若 %AB 与 CoV 同向、φ 与 X5 同向而两组互斥,是否说明测的是两个构造(”腹式化”与”协调化”)?

    方法学要求:预注册的是方向,不是 p 值;每中心同时报方向、效应量与该中心自己的 MDD——方向对但效应量小于本中心 MDD 的,记为”不可判定”,不得计入一致;汇总用方向一致性统计(sign test / 方向 κ),不直接合并效应量。

    相关阅读

    证据等级: C
    参考文献: PMID 9387956, PMID 41091340, PMID 21867540, PMID 25236906, PMID 11714414, PMID 37274016, PMID 7091869, PMID 11181615, PMID 9210815, PMID 28767680, PMID 26387118, PMID 20335298

    L1 网络状态:「该主题尚未进入 L1 对应网络」——CORRESPONDENCE_NETWORK.md 节点表仅把「相位/节律」列为物理侧节点(备注「周期性耦合的关键」),把「呼吸率/呼吸模式」列为生理侧观测项(备注「与节律耦合最直接」),二者均未与易学侧任何节点连成边;边表 3.1–3.4 只登记五行↔五脏/五色/五音/五方/五季/五志、八卦↔五行、天干地支↔五行,无「丹田呼吸—胸腹相位角—三模态」相关边。EDGE_TABLE_science-yixue.md 仅有概念计数(丹田 51 篇 / 气功 48 篇 / 呼吸 189 篇),无方向一致性边。故本文封顶 C。


    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。

    No One Has Verified It: Cross-Center Directional Consistency of Dan Tian Breathing When All Three Modalities Shift in the Same Direction

    Evidence level: C

    The value of the triangular verification does not lie in all three modalities nodding; it lies in each one being able to independently shake its head.

    Main Text

    The previous article assigned “universal rulers” to %AB, φ, X5, and CoV—specifically ICC thresholds and MDD magnitudes. A ruler may be accurate, but that does not mean the measured direction is correct. This constitutes two entirely independent falsification pathways. An instrument can consistently overestimate by 10% while maintaining an ICC of 0.9; it measures the same bias across all variables, remaining consistent yet systematically deviating from the true value.

    This is particularly dangerous in Dantian parameters due to their inherently small effect sizes. Recent human studies on EIT measuring tidal volume provided a precise warning: while the within-subject impedance-volume regression showed an average $R^2$ of 0.89 for 27 subjects (appearing robust), the pooled population $R^2$ dropped to 0.485 (PMID 41091340). Individual alignment does not may support population alignment—stepping on this exact trap occurs when merging data across centers. Furthermore, bias is often not constant but scales with increasing values (proportional bias); on parameters where changes are merely a few percentage points, this can masquerade as a “response.”

    Secondly, the assumption that three modalities shift in the same direction has already been broken by mature literature. A classic comparison between preterm and term infants (PMID 9387956) showed phase angles of 60.6° vs. 12.5° ($p<0.0001$) and breathing work indices of 1.35 vs. 1.01 ($p=0.001$), both differing significantly. However, within the same dataset, the thoracic contribution ratio showed 25.5% vs. 36.3%, $p=0.11$, with no significant difference. The same construct, the same subjects, and the same expected direction of change—%AB remained static. Therefore, “three modalities should shift in the same direction” is not a default premise but a hypothesis requiring individual verification.

    Thirdly, the threshold for cross-modal consistency must be estimated via “cross-method” standards, not “repeated measures of the same method.” Hierarchies of measurement invariance clearly state that after structural consistency (configural) and metric consistency, one must pass scalar consistency before comparing means (PMID 26387118). Many indicators stall at the first two levels: strong correlations and stable rankings exist, yet different intercepts mean the means are not measured on the same ruler. This is likely the most probable scenario for the Dantian triad—all shifting toward “better,” but incomparable regarding “how much better.”

    Fourth, center effects represent structured noise rather than randomness. Randomized controlled trial data from 237 centers and 9,978 cases showed that while the inter-center outcome OR ranged widely (0.27–3.71, a thirteen-fold difference), the treatment effect remained nearly unchanged (1.22→1.24, PMID 21867540). This result is double-edged: the positive aspect is that “large inter-center baseline differences” do not necessarily flip effect directions; the negative aspect is it tempts researchers to conclude “centers are irrelevant,” thereby skipping directional testing entirely. What must be tested is the “effect × center” interaction, not the “baseline × center” difference.

    Fifth, direction reversal has a formal methodological name: the alternative endpoint paradox—where a marker correlates positively with an outcome, yet the intervention produces opposite effects on both (PMID 25236906). Real-world costs are documented: antiarrhythmic drugs effectively reduce ventricular premature beats but significantly increase mortality (PMID 11714414). More subtly, switching the same dataset to “absolute change” or “proportion of responders” endpoints can lead to conflicting conclusions, with extreme cutoff values most prone to generating conflict (PMID 37274016).

    Sixth, consistency holds under quiet conditions. RIP and photoplethysmography measured thoraco-abdominal asynchrony consistently within Bland-Altman limits under both rest and motion (PMID 28767680)—positive evidence. However, with 10 adults performing natural breathing, motion artifacts, simulated airway obstruction, snoring, and coughing: no single sensor performed optimally under all actions; inter-sensor correlations were high during natural breathing but collapsed upon entering artifact segments (PMID 9210815). RIP calibration is also fragile: after changing body positions, equal-volume maneuver calibration yielded only 38%–65% of measurements within ±10% (PMID 7091869); while uncalibrated QDC methods provide correct coefficients only when tidal volume is constant or approximately constant under mathematical validation, results under spontaneous breathing heavily depend on segment selection criteria (PMID 11181615). The practice state is precisely the side where “conditions change”: slower breathing, larger amplitudes, and active abdominal wall engagement.

    Finally, the “direction” of the three modalities is not inherently the same. %AB is a relative measure; its denominator changes, moving the value. φ is an angle crossing zero. CoV represents spatial distribution. X5 strongly depends on lung volume; slow deep breathing itself alters the volume trajectory (PMID 20335298). X5 changes as a secondary effect of volume, not independent evidence of “qi sinking.”

    Therefore, the sample unit for this test is the center × modality cell, not the subject. With 3 centers × 4 modalities, there are only 12 cells; a single direction flip is sufficient to transform “all same direction” into “7/12.” Calculate whether this grid size is sufficient for directional voting before collecting data—this saves far more than post-hoc explanations of why a specific center differs.

    Evidence and Limitations

    Verified: Different parameters under the same construct can yield opposite conclusions (PMID 9387956); high within-subject correlation does not may support population utility (PMID 41091340); large inter-center baseline heterogeneity does not necessarily alter effect direction (PMID 21867540); alternative endpoints can reverse directions and cause actual harm (PMID 25236906, PMID 11714414); different definitions of the same endpoint can cause conflicting conclusions (PMID 37274016); RIP calibration is highly sensitive to body position and breathing conditions (PMID 7091869, PMID 11181615); sensor consistency degrades under artifact conditions (PMID 9210815).

    Partially Verified: RIP and OEP show good consistency under rest and motion (PMID 28767680), indicating comparability between methods can hold under controlled conditions; the measurement invariance framework provides a decision ladder for “when means can be compared” (PMID 26387118).

    To Be Verified: ① Do %AB/φ/X5/CoV effect directions share the same sign after rotating ≥3 centers among the same practitioners? ② Is the consistency rate significantly higher than random (directional voting test)? ③ Can protocol variables (body position, calibration method, positioning aid, acquisition duration) explain direction reversals? ④ If %AB and CoV shift in the same direction while φ and X5 shift in the same direction but the two groups are mutually exclusive, does this indicate measurement of two constructs (“abdominalization” vs. “coordination”)?

    Methodological Requirements: Pre-registration must focus on direction, not p-values; report direction, effect size, and that center’s specific MDD simultaneously for each center. Record as “indeterminate” if direction aligns but effect size is smaller than the center’s MDD; do not count this as consistent. Aggregate using directional consistency statistics (sign test / directional κ), not by directly merging effect sizes.

    Related Reading

    Evidence Level: C
    References: PMID 9387956, PMID 41091340, PMID 21867540, PMID 25236906, PMID 11714414, PMID 37274016, PMID 7091869, PMID 11181615, PMID 9210815, PMID 28767680, PMID 26387118, PMID 20335298

    L1 Network Status: “This topic has not yet entered the L1 corresponding network”—The CORRESPONDENCE_NETWORK.md node table lists only “Phase/Rhythm” as a physical-side node (noted “Key for periodic coupling”) and “Breathing Rate/Breathing Pattern” as a physiological-side observation item (noted “Most directly coupled with rhythm”); neither connects to any Yijing Studies node via an edge. Edge Tables 3.1–3.4 only register Five Phases ↔ Five Organs/Five Colors/Five Tones/Five Directions/Five Seasons/Five Emotions, Eight Trigrams ↔ Five Phases, Heavenly Stems and Earthly Branches ↔ Five Phases; no edges exist regarding “Dantian Respiration—Thoraco-Abdominal Phase Angle—Three Modalities.” EDGE_TABLE_science-yixue.md contains only concept counts (Dantian: 51 papers / Qigong: 48 papers / Breathing: 189 papers), no directional consistency edges. Thus, this article is capped at Level C.


    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.

  • 「练和不练一个样」为什么不能当功夫上身的证据:两态 φ 差值的等效性边界检验

    Evidence level: C

    资深者练功态与静息态的胸腹相位角之差接近零,有三种解释——功夫内化了、他根本没进入练功态、或者这个差值本身测不准。三种在当前数据上长得一模一样。

    第一坑:把「测不出差别」当成「没有差别」

    两组一比,p>0.05,就宣布「趋同成立」。这不成立:不显著只说明数据在零假设下不奇怪,不排除「真有差别但样本太小、变异太大,看不见」(PMID 39157984)。

    要主张「足够小」,得预先定一个最小感兴趣效应量,做等效性检验,证明差值置信区间整体落在边界内(PMID 30873486)。没有方法能证明差值恰好为零,只能证明它小于你事先认账的那个界。

    界怎么定才是关键。5° 以内算「一样」?15° 才算?这个数必须在看数据前定,且要有生理依据——比如按 φ 的日间测量误差来定。定完再看结果,等效检验就只是给「我早知道」盖章。

    第二坑:差值是最不可靠的量

    Δφ = φ练功 − φ静息 是差值分数,信度天然低于两个成分各自的信度,成分越相关塌得越狠。n=139 的 fMRI 研究里,两条件各自内部一致性极好(SB>0.94),两者之差却只有 SB=−0.06,因为两条件相关高达 r=0.97(PMID 29458188)。两个几乎一样的东西相减,剩下的基本是噪声。

    呼吸这边同理:OEP 相位角日间 ICC 只有 0.4–0.7,时间比参数日间根本不可靠;而胸腹容积贡献比的日间 ICC 可超 0.9(PMID 40732591)。

    意思很直白:容积贡献比适合做纵向终点,相位角不适合。 拿全场最不稳的数去算一个更不稳的差值,做之前必须先报 Δφ 自己的重测 ICC 和最小可检测变化。若这个值本身就大于要判定的等效边界,设计在数学上就不可能给出答案。

    第三坑:自动性不等于「两态一样」

    运动学习里,「自动化」的操作定义从来不是「有提示和没提示一样」,而是占用注意资源变少——序列反应任务练到一定程度,加一个数音的次级任务,反应时不再被拖慢(PMID 36871080)。

    但代价下降也不等于不花资源。网球专家双任务下认知成绩掉得比中级选手少、回球准确性保持稳定,作者的解释是任务优先策略(PMID 35051842)。

    更根本的是:自动行为靠情境线索触发,不会永远在线。n=533 研究把自动性与常规性分成两个独立因子,指出自动行为由情境触发、不带目的(PMID 31680711);n=1244 追踪显示自动性与常规、情绪线索一致性相关(PMID 25864705)。

    搬到丹田呼吸上:功夫上身的合理描述是「一坐下、一调息,进入练功态更快更省更抗干扰」,不是「两态一样」。 前者要求两态可区分但切换便宜,后者要求两态等价——相反的预测。

    还有一层:你越说「随便呼吸」,他越不随便

    静息态本身被污染。74 人研究显示:只要受试者知道呼吸在被记录,吸气时间、呼气时间都延长,叹息与呼气末停顿几乎消失,呼吸被「规整化」(PMID 9032510)。82 人实验更直接:知道有人在数呼吸,呼吸率平均低 2.13 次/分(PMID 28426897)。

    杀伤在于:资深者更清楚「什么样子算练得好」,静息态更易被自我规整,Δφ 被人为压小——恰好压向「趋同成立」。这是系统偏倚,加样本量消不掉。

    反向力量也存在:一次实时相位角反馈干预就能把异常呼吸模式的 φ 拉向正常(PMID 34073590)。φ 高度可被即时移动,既是它当干预终点的优点,也是它当稳定特质指标的致命伤。

    现有证据其实指向别处

    若「练久了静息态就变成练功态」,静息指标应有年限梯度。数据只支持一半:4 周装置引导慢呼吸,诊所静息血压与呼吸率下降,但 24 小时血压和夜间呼吸模式没变(PMID 20200548);重度 COPD 患者高强度训练后呼吸模式改变,12 个月未维持(PMID 14718117);31 名长年冥想者对比 38 名对照,静息呼吸率平均低 1.6 次/分,但组内这差异只与闭关累计时长相关(闭关翻倍降 0.7 次/分),与日常练习时长无关(PMID 27272738);呼吸参数的个体排序确实稳定,6 周重测相关 0.7–0.8(PMID 16817607)。

    拼起来:长期练习可能在静息态留痕,但载体更可能是呼吸率与容积、不是相位角;驱动它的是高强度集中练习,不是年限本身。用「年限 × Δφ 递减」做主线,大概率得到又小又不稳的系数。

    怎么才算数:四格判据

    一、Δφ 与 %AB 的 Δ 双双落在等效界内,且两者最小可检测变化都小于等效界 → 趋同成立。此时才谈「内化」,且必须同时报「资深者还能否按指令进一步改变 φ」——连指令都推不动不是内化,是功能丧失。

    二、Δφ 趋零但 %AB 的 Δ 仍显著 → 指标选错,结论改口,此前基于 φ 差值的组间比较全部降级为探索性。

    三、Δφ 不趋零,但进入练功态的潜伏期随年限缩短、双任务干扰随年限减小 → 这才是自动性的正确读法,原假设方向要改。

    四、Δφ 趋零,但「被告知在测」与「不知情」两条件下的 Δφ 差异随年限放大 → 趋同是自我规整的产物。这格最容易漏,也最可能是真相。

    证据与局限

    已验证:非显著不能当无差异(39157984、30873486);差值分数在高相关成分间信度塌陷(29458188);OEP 相位角日间信度低于容积贡献比(40732591);知晓被测量会系统性改变呼吸(9032510、28426897);自动化指注意代价下降而非两态等价(36871080、35051842);自动行为依赖情境线索(31680711、25864705)。

    部分验证:长期练习在静息呼吸留痕,但驱动因素是集中闭关而非日常年限(27272738);短程干预改变不能长期维持(20200548、14718117);φ 可被单次反馈移动(34073590);呼吸参数个体排序稳定(16817607)。

    待验证:核心假设「两态 φ 差值随年限单调收窄」无任何直接数据。该人群 Δφ 的重测信度、最小可检测变化、等效边界全部未知;「资深者静息态被自我规整更多」也未实测。在 Δφ 的最小可检测变化被测出之前,这个假设不具备可检验性。

    L1 网络状态:该主题尚未进入 L1 对应网络。CORRESPONDENCE_NETWORK.md 生理/体验侧节点表虽列有「呼吸率 / 呼吸模式」与「气感报告」观测项,但均未与易学或中医节点成边;「相位 / 节律」仅作物理侧节点存在。EDGE_TABLE_science-yixue.md 仅有「丹田 51 篇 / 气功 48 篇 / 呼吸 189 篇」文章计数,无层级与状态判定。故证据等级最高只能标 C,文中机制表述均为待检验假说,不构成已验证机制。

    相关阅读

    证据等级: C
    参考文献: PMID 39157984, PMID 30873486, PMID 29458188, PMID 40732591, PMID 9032510, PMID 28426897, PMID 36871080, PMID 35051842, PMID 31680711, PMID 25864705, PMID 27272738, PMID 20200548, PMID 14718117, PMID 34073590, PMID 16817607


    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。

    Why \”practicing and not practicing are the same\” cannot serve as evidence of martial arts mastery: Equivalence boundary testing for the two-state phi difference

    Evidence level: C

    Experts report that the difference in thoraco-abdominal phase angles between the practiced state and resting state approaches zero. There are three explanations: the skill has internalized, they never entered the practiced state, or the difference itself is unmeasurable. All three yield identical results under current data.

    Pitfall One: Mistaking “Failure to Detect a Difference” for “No Difference”

    Comparing two groups yields $p > 0.05$, leading to the declaration of “convergence.” This is invalid: non-significance only indicates that data are not unusual under the null hypothesis; it does not exclude “there is a true difference but the sample size is too small or variability too large to see it” (PMID 39157984).

    To claim “sufficiently small,” one must pre-specify a minimum effect size of interest and conduct an equivalence test, proving the confidence interval for the difference falls entirely within the boundary (PMID 30873486). No method can prove the difference is exactly zero; one can only prove it is smaller than a boundary you acknowledge beforehand.

    How the boundary is defined is critical. Is within 5° considered “the same”? Is 15° required? This number must be set before viewing the data and grounded in physiological rationale—for example, based on daily measurement error for $\phi$. Setting it after seeing results turns the equivalence test into merely stamping a seal on “what I already knew.”

    Pitfall Two: The Difference is the Least Reliable Metric

    $\Delta\phi = \phi_{\text{practiced}} – \phi_{\text{resting}}$ is a difference score, which naturally has lower reliability than the reliability of either component alone; reliability collapses further as components become more correlated. In an $n=139$ fMRI study, internal consistency for both conditions was excellent ($SB > 0.94$), yet their difference yielded only $SB = -0.06$ because the correlation between conditions was as high as $r=0.97$ (PMID 29458188). Subtracting two nearly identical things leaves mostly noise.

    The same applies to breathing: OEP phase angles have a daily ICC of only 0.4–0.7, and time-ratio parameters are fundamentally unreliable on a daily basis; conversely, the thoraco-abdominal volume contribution ratio can exceed an ICC of 0.9 daily (PMID 40732591).

    The implication is straightforward: Volume contribution ratios are suitable for longitudinal endpoints, while phase angles are not. Using the least stable metric to calculate an even less stable difference requires reporting $\Delta\phi$’s own retest ICC and minimum detectable change beforehand. If this value itself exceeds the equivalence boundary intended for judgment, the design is mathematically incapable of providing an answer.

    Pitfall Three: Automaticity Does Not Equal “Two States Are the Same”

    In motor learning, the operational definition of “automation” is never “the same with and without cues,” but rather a reduction in attentional resource allocation—where sequence reaction tasks reach a threshold where adding a secondary tone task no longer slows reaction time (PMID 36871080).

    However, cost reduction does not equate to zero resource expenditure. Under dual-task conditions, tennis experts show smaller declines in cognitive performance and stable ball accuracy compared to intermediate players; authors attribute this to a task-priority strategy (PMID 35051842).

    More fundamentally: automatic behaviors are triggered by contextual cues and do not remain perpetually online. An $n=533$ study separated automaticity from routine into two independent factors, indicating automatic behaviors are context-triggered and goalless (PMID 31680711); an $n=1244$ longitudinal tracking showed automaticity correlates with routine and emotional cues (PMID 25864705).

    Applied to Dantian breathing: A reasonable description of “skill acquisition” is “faster, more economical, and more resistant to interference upon sitting and adjusting breath,” not “two states are identical.” The former requires distinguishable states with cheap switching; the latter requires equivalent states—opposing predictions.

    Another Layer: The More You Say “Just Breathe,” the Less It Becomes Random

    The resting state itself is contaminated. A study of 74 subjects showed: as soon as subjects knew breathing was being recorded, inspiratory and expiratory times extended, sighs and expiratory pauses almost vanished, and breathing became “regularized” (PMID 9032510). An 82-person experiment was even more direct: knowing someone was counting breaths lowered the breathing rate by an average of 2.13 breaths/min (PMID 28426897).

    The damage lies in: Experts are more aware of “what constitutes good practice,” making the resting state more susceptible to self-regulation, thereby artificially compressing $\Delta\phi$—exactly toward “convergence established.” This is a systematic bias that sample size cannot eliminate.

    The reverse force also exists: a single real-time phase angle feedback intervention can pull abnormal breathing pattern $\phi$ values toward normal (PMID 34073590). $\phi$ is highly mobile via immediate adjustment; this is its advantage as an intervention endpoint, but also its fatal flaw as a stable trait indicator.

    Existing Evidence Points Elsewhere

    If “practicing long-term transforms the resting state into the practiced state,” resting indicators should show a gradient over years. Data support only half: 4 weeks of device-guided slow breathing lowered clinic resting blood pressure and breathing rate, but 24-hour blood pressure and nocturnal breathing patterns remained unchanged (PMID 20200548); severe COPD patients showed breathing pattern changes after high-intensity training, but these were not maintained at 12 months (PMID 14718117); comparing 31 long-term meditators to 38 controls showed resting breathing rates averaged 1.6 breaths/min lower, but within-group differences correlated only with cumulative retreat duration (doubling retreat reduced rate by 0.7 breaths/min), unrelated to daily practice duration (PMID 27272738); individual ranking of breathing parameters is indeed stable, with 6-week retest correlations of 0.7–0.8 (PMID 16817607).

    Put together: long-term practice may leave traces in the resting state, but the carrier is likely breathing rate and volume rather than phase angles; the driver is high-intensity focused practice, not years themselves. Using “years $\times$ $\Delta\phi$ decay” as the main thread will likely yield small and unstable coefficients.

    How to Count: The Four-Grid Criteria

    1. Both $\Delta\phi$ and $\Delta \%AB$ fall within equivalence boundaries, AND both minimum detectable changes are smaller than the equivalence boundary $\rightarrow$ Convergence established. Only then discuss “internalization,” and must simultaneously report whether experts can further alter $\phi$ via instruction—failure to move even with instruction is not internalization but functional loss.

    2. $\Delta\phi$ approaches zero but $\Delta \%AB$ remains significant $\rightarrow$ Wrong metric selected; conclusions must be revised; all between-group comparisons based on $\phi$ differences downgrade to exploratory.

    3. $\Delta\phi$ does not approach zero, but latency to enter practiced state shortens with years and dual-task interference decreases with years $\rightarrow$ This is the correct reading of automaticity; original hypothesis direction must be revised.

    4. $\Delta\phi$ approaches zero, but the difference in $\Delta\phi$ between “informed” and “uninformed” conditions amplifies with years $\rightarrow$ Convergence is a product of self-regulation. This category is most easily missed and likely the truth.

    Evidence and Limitations

    Verified: Non-significance cannot serve as no-difference (39157984, 30873486); difference scores collapse in reliability between highly correlated components (29458188); OEP phase angles have lower daily reliability than volume contribution ratios (40732591); knowing measurement alters breathing systematically (9032510, 28426897); automation refers to reduced attentional cost, not two-state equivalence (36871080, 35051842); automatic behaviors depend on contextual cues (31680711, 25864705).

    Partially Verified: Long-term practice leaves traces in resting breathing, but drivers are focused retreats rather than daily years (27272738); short-term interventions change but do not maintain long-term (20200548, 14718117); $\phi$ can be moved by single feedback (34073590); individual ranking of breathing parameters is stable (16817607).

    To Be Verified: Core hypothesis “two-state $\phi$ difference monotonically narrows with years” has no direct data. Retest reliability, minimum detectable change, and equivalence boundaries for this population’s $\Delta\phi$ are all unknown; “expert resting state is more self-regulated” has not been measured. Before $\Delta\phi$’s minimum detectable change is measured, this hypothesis lacks testability.

    L1 Network Status: This topic has not yet entered the L1 corresponding network. While CORRESPONDENCE_NETWORK.md lists physiological/experiential nodes such as “breathing rate / breathing patterns” and “qi sensation reports,” none form edges with Yijing or TCM nodes; “phase / rhythm” exists only as a physical-side node. EDGE_TABLE_science-yixue.md contains only article counts (“Dantian 51 / Qigong 48 / Breathing 189”), lacking hierarchy or state determination. Thus, the highest evidence level is C; all mechanistic descriptions in this text remain unverified hypotheses and do not constitute verified mechanisms.

    Related Reading

    Evidence Level: C
    References: PMID 39157984, PMID 30873486, PMID 29458188, PMID 40732591, PMID 9032510, PMID 28426897, PMID 36871080, PMID 35051842, PMID 31680711, PMID 25864705, PMID 27272738, PMID 20200548, PMID 14718117, PMID 34073590, PMID 16817607


    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.

  • 功夫「上身」的两只表:φ 轨迹拐点与气感首次报告,为什么不能直接对表

    Evidence level: C

    一边是仪器画出的胸腹相位角曲线,一边是练功日记里”第一次感觉到气”的那一天。把两表一叠,看起来就给了”功夫上身”一个客观锚点。但这个叠加里有两个独立的坑:拐点估不准、日记记不准,而气感本身还会被期望提前。

    正文

    上一篇提出了丹田呼吸的 φ(胸腹相位角)倒 U 轨迹假设:初期刻意用力使胸腹异步升高,技能内化后回落。下一个问题自然是——拐点何时发生,能不能和”气感第一次出现”对上? 直觉极顺:曲线一拐、人一感觉到,玄妙体验就装上了仪表。可惜把两表直接叠起来,方法学上很难站住。

    第一只表:拐点估不准。 变点估计精度对样本量和拐点位置极度敏感。2025 年一项模拟研究比较五种方法,结论很干脆:样本量仅 100 时所有方法都表现糟糕——偏差大、RMSE 大、95% 区间覆盖率远低于名义水平;即便样本量上去,只要真实拐点靠近观测边界,估计依然失效,甚至给出超出观测范围的点估计(PMID 40585262)。而”入练后 0、1、3、6、12 个月测五次”只有 4 段区间,拐点无论落哪儿都离边界不远。更要命的是:拐点置信区间宽度很可能超过整个研究期跨度。这时”拐点在 3 个月”与”6 个月”统计上分不开,而气感报告差一个月就足以让同窗对齐的结论翻转。

    第二只表:日记记不准。 气感”首次出现时间”几乎总是回溯数据,而回溯日期有系统性错位,学名叫 telescoping——事件在记忆里被挪近或挪远,且不只发生在临床人群:前瞻研究里同一批孩子的同一个最早记忆,一年后复述时日期被系统性后置(PMID 24588518);用重复横断面校正发现,25 岁回溯的首次饮酒年龄平均比校正估计晚了近两年(PMID 21544259)。练功日记还叠一层:依从性随时间下降且有规律。一项 9 周生态瞬时评估研究显示,早间定时评估依从性从 86% 起步、每周掉 2%,随机评估从 58% 起步、每周掉 1%(PMID 37347538);早期依从性对全程依从性有很强预测力(PMID 34541425),青少年样本同样随时程下降、负性情绪高的日子更低(PMID 28080107)。于是”气感首次报告”有个系统性偏晚的机制:日记掉得越厉害,该被记下的那天越可能漏掉,最后记下的”第一次”其实是第二、第三次。

    第三只表:气感会被期望提前。 期望效应在躯体症状领域证据清晰。一项跨症状综述指出,对呼吸困难和瘙痒,期望能预测安慰剂/反安慰剂效应强度,但心率、肺功能、胃活动这类生理指标很少同步改变(PMID 31312148)——报告动了、生理没动。气功领域也支持这种分层:早期综述记录了练功中的温热、寒凉、麻木、酸胀、飘浮、身体形象放大或缩小等体验,且随练习终止而消失(PMID 8088187);一项八段锦研究更显式展示了主观与生理的不同向——练后主观平静、身体激活感、专注度、主观活力全套升高,生理层面却是副交感张力下降、心率上升(PMID 34211574)。

    再细一层:人对自己的学习曲线能猜对形状、猜错参数。六预注册实验(n=500)发现,参与者能正确预判技能学习呈指数衰减,但对起点水平、平均表现、改善速率系统性估错(PMID 39978146)。放进练功场景:新手大概都知道”三个月会有点感觉”,但不知道自己的速度。当期望自带时间标签,气感报告的时间点就不再是独立的第三观测,而可能是被期望设定好的。 于是”气感出现在第 3 个月”与”φ 拐点在第 3 个月”的重合,可能只是”我以为三个月该有了”与”我恰好测在三个月”的重合。

    还有一条方向性问题:测量本身会改变被测对象。刻意做膈式呼吸反而使相位角显著升高(PMID 7697243),而实时相位角反馈能在一次干预内把相位角拉向零(PMID 34073590)。φ 在被”认真练”和”被显示出来”时会动——日志越密、聚焦越强,拐点就越不像自然成熟的拐点。

    所以真正的产出不是”对上了/没对上”,而是一套判据。 四条互斥结局:① 气感首次持续报告早于拐点超过一个随访间隔 → 气感是先行指标或被期望驱动;② 拐点早于气感超过一个间隔 → 身体先变、感知滞后;③ 落在同一窗口 → 才谈得上锚定,但仍须排除共同驱动(累积练习量、季节、带教更换、同期生活事件);④ 个体轨迹检不出拐点、只有组平均才有 → 倒 U 在个体层面不成立,只能降级为群体描述。

    配套三条方法学要求:用”首次持续报告”(连续≥2 次随访都报告)代替”首次报告”,滤掉一次性模糊体感;日记改成每周固定日、带日历锚点、问绝对日期,压缩 telescoping 空间;预注册拐点搜索窗与最少随访点数,并同时报告拐点 CI 与”可分辨的最小间隔”——若 CI 宽于一个随访间隔,应直接宣布该设计的对齐检验不可行。

    更根本的一层:自我报告与客观测量之间的差,本身已是成熟的量化构念。内感受研究把它定义为”特质预测误差”——客观心跳检测准确度与主观躯体敏感度之差,该差在自闭特质人群中与情绪识别缺陷和焦虑相关(PMID 26724504),在功能性发作患者中能预测解离程度与发作频率(PMID 32501547)。启发是:“气感评分 − φ 偏离度”这个差,可能比”气感是否与 φ 同步”更有信息量——它测的正是”这个人对自己身体状态的读数有多准”,而这恰恰可训练。

    证据与局限

    已验证: 变点估计在样本量小或拐点靠近观测边界时偏差大、区间覆盖率不足(PMID 40585262);回溯日期存在系统性 telescoping 偏差,前瞻与校正数据可量化其幅度(PMID 24588518、21544259);长时程自我报告依从性随时间下降、早期依从性预测全程依从性(PMID 37347538、34541425、28080107);期望可预测呼吸困难与瘙痒的主观强度而生理指标很少同步改变(PMID 31312148);气功练习中的体感随练习终止消失(PMID 8088187);八段锦后主观改善与副交感张力下降同时出现(PMID 34211574);人对学习曲线形状判断准确而参数判断系统性错误(PMID 39978146);自评内感受准确度与客观准确度可分离且差值有独立预测力(PMID 26724504、32501547、34929353);刻意膈式呼吸与实时相位角反馈均可急性改变 φ(PMID 7697243、34073590)。

    部分验证: 上述各条分别成立,但没有任何一项研究把”φ 纵向轨迹拐点”与”气感首次报告时刻”放进同一批受试者、同一时间轴做事件时间配对。”拐点与气感对齐”这一命题本身无任何直接数据,属研究设计提案。

    待验证: ① 新手 5 个随访点的 φ 轨迹能否稳定检出拐点(CI 宽度 vs 一个随访间隔);② “首次持续报告”与”首次报告”的时间差有多大、是否随依从性下降放大;③ 气感报告是否被”预期三个月”这类时间锚定型期望提前(可随机化不同期望脚本检验);④ “气感 − φ 偏离”差值是否随修炼年限收窄(读数变准假说);⑤ 个体层面倒 U 是否真实,还是组平均的聚合假象。

    相关阅读

    L1 网络状态:该主题尚未进入 L1 对应网络。CORRESPONDENCE_NETWORK.md 的边表中不存在以「气感报告」或「技能习得时间进程」为源或目标的边——生理/体验侧节点表虽列有「气感报告(部位/强度/时序/热麻胀沉浮)」观测项,但未与任何易学或生理节点连边;EDGE_TABLE_science-yixue.md 仅有「丹田 51 篇 / 气功 48 篇 / 呼吸 189 篇」的文章计数,无层级与状态判定。故证据等级按硬规则封顶为 C。
    证据等级: C(各条方法学证据均为 PubMed 直接支撑,但”φ 拐点与气感报告事件时间对齐”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,故整体降为 C——本文是对一个测量假设的可行性审查,不是已验证机制)
    参考文献: PMID 40585262, PMID 24588518, PMID 21544259, PMID 37347538, PMID 34541425, PMID 28080107, PMID 31312148, PMID 8088187, PMID 34211574, PMID 39978146, PMID 26724504, PMID 32501547, PMID 34929353, PMID 7697243, PMID 34073590


    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。

    Two Watches with \”Kung Fu\” Inside: φ Trajectory Inflection Points and First Report of Qi Sensation, Why Can’t We Directly Compare Them

    Evidence level: C

    On one side is the chest-abdominal phase angle curve drawn by an instrument; on the other is the day recorded in a practice diary as “the first time one felt qi.” Superimposing these two tables appears to provide an objective anchor for “skill embodiment.” However, this superposition contains two distinct pitfalls: inflection point estimation is imprecise, diary recording is unreliable, and qi sensations themselves are prone to premature expectation effects.

    Main Text

    The previous article proposed the $\phi$ (chest-abdominal phase angle) inverted U-shaped trajectory hypothesis for dantian breathing: initial deliberate effort causes asynchronous elevation of the chest and abdomen, followed by a return to coordination upon skill internalization. The next logical question is: When does the inflection point occur, and can it align with “the first appearance of qi sensation”? The intuition is compelling: once the curve bends and a person feels it, the esoteric experience is effectively mounted onto an instrument. Unfortunately, directly superimposing the two tables is methodologically unsound.

    Table 1: Inflection Point Estimation is Imprecise. Change-point estimation precision is highly sensitive to sample size and the location of the inflection point. A 2025 simulation study comparing five methods concluded unequivocally that with a sample size of only 100, all methods performed poorly—exhibiting large bias, large RMSE, and 95% interval coverage rates far below nominal levels; even with larger samples, estimation fails if the true inflection point is near an observation boundary, sometimes yielding point estimates outside the observed range (PMID 40585262). In a typical practice regimen measuring five times at 0, 1, 3, 6, and 12 months post-initiation, there are only four intervals; regardless of where the inflection point falls, it remains close to a boundary. More critically: the width of the inflection point confidence interval may exceed the entire study duration. At this stage, an inflection at “3 months” versus “6 months” is statistically indistinguishable, yet a one-month discrepancy in qi sensation reports can flip aligned conclusions among peers.

    Table 2: Diary Recording is Unreliable. The “first appearance time” of qi sensation is almost always retrospective data, and retrospective dates suffer from systematic misalignment known as telescoping—events are shifted closer or further back in memory, a phenomenon not limited to clinical populations: in prospective studies, the same earliest memory from a group of children was systematically postponed by one year when recounted after twelve months (PMID 24588518); using repeated cross-sectional correction, the reported age of first drinking at age 25 averaged nearly two years later than the corrected estimate (PMID 21544259). Practice diaries add another layer: adherence declines systematically over time. A 9-week ecological momentary assessment study showed that adherence for scheduled morning assessments started at 86% and dropped 2% per week, while random assessments started at 58% and dropped 1% per week (PMID 37347538); early adherence strongly predicts overall adherence (PMID 34541425), with similar declines observed in adolescent samples and lower rates on days with negative emotions (PMID 28080107). Consequently, “first reported qi sensation” suffers a systematic mechanism of delay: the more the diary slips, the more likely that day should have been recorded is missed, meaning the “first” finally recorded is actually the second or third.

    Table 3: Qi Sensation May Be Driven by Expectation. Expectation effects are well-documented in somatic symptom domains. A cross-symptom review indicated that expectation predicts the intensity of placebo/nocebo effects for dyspnea and pruritus, but physiological indicators such as heart rate, lung function, and gastric activity rarely change synchronously (PMID 31312148)—reports shift while physiology does not. The qigong domain also supports this stratification: early reviews recorded experiences during practice such as warmth, coldness, numbness, soreness, floating, body image expansion or contraction, all of which disappear upon cessation of practice (PMID 8088187); a Baduanjin study more explicitly demonstrated divergent subjective and physiological trajectories—post-practice subjective calmness, bodily activation, focus, and vitality all increased, while physiological measures showed decreased parasympathetic tone and increased heart rate (PMID 342121574).

    Digging deeper: Individuals can correctly anticipate the shape of their learning curve but err on parameters. Six preregistered experiments ($n=500$) found that participants could accurately predict skill learning followed an exponential decay but systematically misestimated starting levels, average performance, and improvement rates (PMID 39978146). Applied to practice scenarios: novices generally know “some sensation will emerge around three months” but lack knowledge of their individual pace. When expectation carries its own temporal label, the reported time point of qi sensation is no longer an independent third observation but may be pre-set by expectation. Thus, the coincidence of “qi sensation appearing in month 3” and “$\phi$ inflection occurring in month 3” may merely reflect the convergence of “I expected it by month 3” and “I happened to measure in month 3.”

    Another directional issue: measurement itself alters the measured subject. Deliberate diaphragmatic breathing actually significantly increases phase angle (PMID 7697243), whereas real-time phase angle feedback can pull the phase angle toward zero within a single intervention session (PMID 34073590). $\phi$ shifts when “practiced seriously” and “displayed”—denser logs and stronger focus make the inflection point less like a naturally matured one.

    Therefore, the true output is not “aligned/not aligned,” but a set of criteria. Four mutually exclusive outcomes: ① First sustained report of qi sensation occurs more than one follow-up interval before the inflection point → qi sensation is a leading indicator or driven by expectation; ② Inflection point occurs more than one interval before qi sensation → bodily change precedes perceptual lag; ③ Falls within the same window → alignment is possible, yet common drivers must still be excluded (cumulative practice volume, season, instructor change, concurrent life events); ④ Individual trajectories show no inflection point while group averages do → the inverted U does not hold at the individual level and can only be downgraded to a group description.

    Three methodological requirements accompany this: use “first sustained report” (reported continuously in ≥2 follow-ups) instead of “first report” to filter out one-off vague sensations; change diaries to fixed weekly dates with calendar anchors and ask for absolute dates to compress telescoping space; preregister inflection point search windows and minimum follow-up points, while simultaneously reporting the inflection point CI and the “minimum resolvable interval”—if the CI exceeds one follow-up interval, declare directly that alignment testing is infeasible for that design.

    A more fundamental layer: the difference between self-report and objective measurement is itself a mature quantitative construct. Interoception research defines this as “trait prediction error”—the difference between objective heart rate detection accuracy and subjective somatic sensitivity. This difference correlates with emotion recognition deficits and anxiety in autistic traits (PMID 26724504) and predicts dissociation severity and attack frequency in functional disorder patients (PMID 32501547). The implication is: the difference “qi sensation score − $\phi$ deviation” may hold more information than whether qi sensation synchronizes with $\phi$”—it measures precisely “how accurate this person’s reading of their own bodily state is,” which can indeed be trained.

    Evidence and Limitations

    Verified: Change-point estimation exhibits large bias and insufficient interval coverage with small sample sizes or inflection points near observation boundaries (PMID 40585262); retrospective dates exhibit systematic telescoping bias, with prospective and corrected data quantifying the magnitude (PMID 24588518, 21544259); long-term self-reported adherence declines over time, with early adherence predicting overall adherence (PMID 37347538, 34541425, 28080107); expectation predicts subjective intensity for dyspnea and pruritus but rarely synchronously alters physiological indicators (PMID 31312148); somatic experiences in qigong practice disappear upon cessation (PMID 8088187); post-Baduanjin subjective improvements co-occur with decreased parasympathetic tone (PMID 342121574); individuals accurately judge learning curve shape but systematically err on parameters (PMID 39978146); self-rated interoceptive accuracy can be separated from objective accuracy, with the difference having independent predictive power (PMID 26724504, 32501547, 34929353); deliberate diaphragmatic breathing and real-time phase angle feedback can acutely alter $\phi$ (PMID 7697243, 34073590).

    Partially Verified: Each of the above holds individually, but no study has paired “the longitudinal trajectory inflection point of $\phi$” with “the first reported moment of qi sensation” within the same cohort and on the same timeline. The proposition itself that “inflection point aligns with qi sensation” lacks direct data and constitutes a research design proposal.

    To Be Verified: ① Can stable inflection points be detected in novice $\phi$ trajectories across five follow-up points (CI width vs. one follow-up interval)? ② What is the time difference between “first sustained report” and “first report,” and does it amplify with declining adherence? ③ Is qi sensation reporting advanced by time-anchor expectations like “expected in three months” (verifiable via randomized expectation scripts)? ④ Does the “qi sensation − $\phi$ deviation” difference narrow with years of cultivation (reading accuracy hypothesis)? ⑤ Is the inverted U true at the individual level, or merely an aggregation artifact of group averages.

    Related Reading

    L1 Network Status: This topic has not yet entered the L1 corresponding network. The edge table in CORRESPONDENCE_NETWORK.md contains no edges originating from or targeting “qi sensation reports” or “skill acquisition time course.” While the physiological/experience node table lists observation items such as “qi sensation reports (location/intensity/temporal pattern/heat-numbness-swelling-floating),” no edges connect these to any Yijing or physiological nodes; EDGE_TABLE_science-yixue.md only contains article counts (“Dantian 51 articles / Qigong 48 articles / Breathing 189 articles”) without hierarchical or status determination. Therefore, the evidence level is capped at C per hard rules.
    Evidence Level: C (Each methodological evidence item is directly supported by PubMed data, but the proposition itself regarding “alignment between $\phi$ inflection points and qi sensation report timing” lacks direct research, and this topic has no corresponding edges in the L1 network; thus, the overall level is downgraded to C—this article reviews the feasibility of a measurement hypothesis rather than verifying an established mechanism)
    References: PMID 40585262, PMID 24588518, PMID 21544259, PMID 37347538, PMID 34541425, PMID 28080107, PMID 31312148, PMID 8088187, PMID 342121574, PMID 39978146, PMID 26724504, PMID 32501547, PMID 34929353, PMID 7697243, PMID 34073590


    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.

  • 呼吸稳不稳,不能只看晃得小不小:丹田修炼者胸腹相位角逐次变异的「冻结式稳定」与「功能性变异」裁决

    Evidence level: C

    一个”晃得小”的呼吸,可能是练到家了,也可能是根本动不了。同一个数字,两种相反的解释。

    正文

    一、这个数字本身没有方向

    胸腹相位角 φ 衡量的是胸腔和腹腔在呼吸周期里谁先动、差多少度。把连续几十次呼吸的 φ 算出一个标准差(φ-SD),直觉上”数越小=越协调越稳”。但这个直觉在两个地方会翻车。

    第一处是算术。 φ 是角度。一次呼吸 179°、下一次 −179°,实际只差 2°,算术标准差却算成 358°。而这一点恰好系统性地坑掉真正练得好的人——胸腹越同步,φ 越靠近 0°,越容易在正负之间跨零。越是”练到家”,越容易被算出一个假的大变异。 正确做法是用圆形离散度(平均合向量长度 R、圆方差 1−R、von Mises 浓度参数)。选哪个也不随便:圆形统计检验的敏感度与假阳性率差异很大,Watson 检验多数情形最好,Modulation Index 只在耦合比高于 2:1 时占优(PMID 33848619)。

    第二处是门槛。 组间差必须先跨过噪声底。φ 的个体内变异系数在早产儿安静睡眠中达 11%–123%,原作者判定该变异来自胸腹相位本身不稳、而非算法误差(PMID 24716708)。成人分区贡献在运动态重测信度虽高(ICC 0.87–0.93,PMID 35226168),但 2026 年范围综述指出成人呼吸模式测量的信度报告极不统一、绝对容积随校准与体位漂移(PMID 42862511)。在算出 φ-SD 的重测信度之前,”修炼者更稳”连能不能测都还没定。

    二、变异性理论给出的方向,可能和假设相反

    运动控制里有一条几十年的主线,专治”变异小是不是好事”。Bernstein 的答案是:学新动作时中枢神经系统先冻结多余自由度,等基本功稳了再释放。2020 年系统综述的结论是冻结还是释放取决于技能类别与目标的交互,不能一刀切(PMID 32221040);棒球击球的实测清楚显示了从冻结到释放的迁移,且释放后出现的是功能性变异——变异没消失,而是被重新分配到”不影响结果”的维度上(PMID 32670156)。

    放到丹田呼吸上,这意味着一个扎心的可能:新手 φ-SD 小,可能是把胸腹锁成了一块板(冻结);修炼者 φ-SD 大,可能是该松的松了(释放)。 原假设”修炼者 φ-SD 更小”的方向,有可能是整个反的。

    一条独立证据从侧面顶住了这一点。15 名健康人做四种呼吸练习的对照研究发现,膈式呼吸是唯一让胸腹异步性变量显著升高的那一种,同时带来最低的胸廓贡献和最高的腹部贡献(PMID 25590447)。腹式呼吸做得越”标准”,胸腹协调的瞬时波动未必越小。 只盯 φ-SD 的设计,会把这种正常波动读成”协调变差”。

    三、另一条线却说小就是稳,关键在”变异的形态”

    反过来的证据也有。清醒呼吸不规则程度与睡眠呼吸暂停患者六个月后能否坚持 CPAP 显著相关(PMID 23080481)——不规则本身确实携带信息。

    调和这两条线的钥匙,是 2023 年一项 526 人、11 个任务的研究:变异的时间结构(lag-1 自相关、功率谱密度斜率)有很高的个体内跨场次重测性,是稳定特质;但它们在解释注意力状态与自评特质上几乎无效;而更复杂的 ARFIMA 长记忆成分重测性很差(PMID 36618326)。变异的”大小”未必是特质,变异的”形态”才更可能是。 若 φ-SD 自己重测不过关,该换的终点是 φ 序列的自相关结构,而不是在 SD 上继续叠样本量。

    四、四个必须先排掉的混杂

    1. 体型。仰卧位下肥胖组腹区容积占比 41%、正常体重组 31%,肋廓贡献 5.7% 对 31.1%(PMID 37852999)。腹区占比直接决定 φ 的可动范围。
    2. 体位。COPD 患者坐位出现胸廓异步、仰卧反而消失,健康对照两种体位都无分区异步(PMID 23412901)。体位一变,φ 的符号级结论都能翻。
    3. 呼吸频率。潮气量低频功率份额能预测 HRV 低频功率差异最多达 79%,作者主张呼吸变异必须始终纳入解释(PMID 24423137)。φ-SD 与呼吸周期变异高度共线。
    4. 有效样本量。呼吸序列有自相关,独立样本数远小于呼吸次数,须用有效自由度估计才能给出可信显著性(PMID 39846001)。10 分钟记录算出的”SD”,置信区间比想象中宽得多。

    五、设计与四格判据

    ≥90 人(新手 / 1–5 年 / 5 年以上各 30),固定体位与倾角,连续记录 ≥10 分钟静息呼吸,逐次呼吸提取 φ、胸廓与腹区贡献、V_T、呼吸周期。φ 离散度用平均合向量长度 R 与圆方差报告,同时报算术 SD 以便与既往文献对照。混杂按 BMI、年龄、性别匹配,呼吸周期作协变量。

    四格判读:

    • 修炼者 φ-SD 小,且 V_T 与周期变异同等小,φ 均值居中不贴 0° → 整体稳定成立
    • 修炼者 φ-SD 小,但 V_T 或周期变异反而更大 → 冻结式稳定:牺牲一个维度换整体稳,不是功夫深
    • 修炼者 φ-SD 大,且多出的变异与换气效率不相关 → 功能性变异:原假设方向错误,”解放”不等于退化
    • 两组 φ-SD 无差异,但 lag-1 自相关或谱斜率有差异 → 指标选错,终点必须更换,既往 φ-SD 组间比较一律降级为探索性

    L1 网络状态:该主题尚未进入 L1 对应网络——CORRESPONDENCE_NETWORK.md 中无任何以「丹田/气功/胸腹相位角/呼吸变异」为源或目标的边,仅在生理侧节点表列有「呼吸率 / 呼吸模式」观测项(备注”与节律耦合最直接”)与「胃肠蠕动」作为「气沉丹田」的候选物;EDGE_TABLE_science-yixue.md 只有「丹田 51 篇 / 气功 48 篇 / 呼吸 189 篇」的文章计数,无层级(L1/L2/L3)与状态(active/open/conflict)判定。按硬约束,本文证据等级封顶 C。

    证据与局限

    已验证:胸腹相位角可用互相关法稳健测量;胸腹分区贡献在运动态重测信度 ICC 0.87–0.93(PMID 35226168);肥胖者仰卧位腹区占比显著高于正常体重者(PMID 37852999);体位改变可使 COPD 胸廓异步性反向(PMID 23412901);膈式呼吸显著升高健康人胸腹异步性(PMID 25590447);变异的时间结构比变异大小更具特质性(PMID 36618326)。

    部分验证:φ 的个体内变异幅度大且源于生理不稳(PMID 24716708,但样本为早产儿);呼吸变异性与 HRV 低频功率耦合(PMID 24423137);自由度冻结→释放的技能迁移在运动任务中成立(PMID 32221040、32670156),但从未在呼吸任务上验证过;清醒呼吸不规则与临床结局关联(PMID 23080481,单一病种)。

    待验证:修炼者与新手静息 φ-SD 的差异方向本身——全世界无直接研究;φ-SD 的成人重测信度与最小可检测变化;”冻结式稳定”在呼吸系统是否存在;φ 圆形离散度与算术 SD 在跨零情形下的偏差量级。

    根本局限:本文是把运动控制领域的变异理论移植到呼吸协调上,属类比推理,不是直接证据。丹田呼吸的”协调”是否与运动技能共享同一套自由度控制机制,本身未被检验。所有结论在获得 φ-SD 重测信度之前都只是设计提案。

    相关阅读

    证据等级: C
    参考文献: PMID 24716708, PMID 42862511, PMID 35226168, PMID 32221040, PMID 32670156, PMID 36618326, PMID 23080481, PMID 37852999, PMID 23412901, PMID 24423137, PMID 39846001, PMID 25590447, PMID 33848619


    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。

    Stability of Breathing Is Not Determined Solely by Small Movements: A Judgment on the \”Frozen Stability\” and \”Functional Variation\” of Thoraco-Abdominal Phase Rhythms in Practit…

    Evidence level: C

    A “small swing” in breathing may indicate mastery, or it may indicate immobility. The same number yields two opposite interpretations.

    Main Text

    I. The Number Itself Has No Direction

    The thoraco-abdominal phase angle $\phi$ measures which compartment moves first and by how many degrees within the respiratory cycle. Calculating a standard deviation ($\phi$-SD) from dozens of consecutive breaths intuitively suggests “smaller numbers = better coordination and stability.” However, this intuition fails in two critical areas.

    First is arithmetic. $\phi$ is an angle. A breath at $179^\circ$ followed by $-179^\circ$ differs physically by only $2^\circ$, yet the arithmetic standard deviation calculates to $358^\circ$. This systematically penalizes true practitioners—those with greater synchronization have $\phi$ values closer to $0^\circ$, making them more prone to crossing the zero boundary. The more “mastered” the practice, the more likely one is to be calculated as having high variability. The correct approach utilizes circular dispersion metrics (mean resultant length $R$, circular variance $1-R$, von Mises concentration parameter). Selection is not arbitrary: circular statistical tests vary significantly in sensitivity and false-positive rates; the Watson test is generally superior, while the Modulation Index only dominates when coupling ratios exceed 2:1 (PMID 33848619).

    Second is the threshold. Inter-group differences must first exceed the noise floor. The intrasubject coefficient of variation for $\phi$ ranges from 11% to 123% in quiet sleep among preterm infants; the original authors attribute this variability to instability in the thoraco-abdominal phase itself rather than algorithmic error (PMID 24716708). While inter-part contribution reliability in adults during movement states is high (ICC 0.87–0.93, PMID 35226168), a 2026 systematic review notes that reliability reports for adult breathing pattern measurements are highly inconsistent, with absolute volumes drifting based on calibration and posture (PMID 42862511). Before establishing the retest reliability of $\phi$-SD, it is undetermined whether “practitioners are more stable” can even be measured.

    II. The Direction Suggested by Variability Theory May Be Opposite to the Hypothesis

    In motor control, a decades-old mainline addresses whether “low variability is good.” Bernstein’s answer is that the central nervous system first freezes excess degrees of freedom when learning new movements, releasing them only after fundamentals are stabilized. A 2020 systematic review concludes that freezing versus release depends on the interaction between skill category and goal, not a universal rule (PMID 32221040); empirical data on baseball batting clearly demonstrates a migration from freezing to release, where the resulting variability is functional variation—the variability does not disappear but is reallocated to dimensions that do not affect the outcome (PMID 32670156).

    Applied to Dantian breathing, this implies a startling possibility: Novices with small $\phi$-SD may be locking the thorax and abdomen into a rigid plate (freezing); practitioners with large $\phi$-SD may be releasing where they should (release). The original hypothesis that “practitioners have smaller $\phi$-SD” may be entirely reversed.

    Independent evidence supports this from the side. A comparative study of 15 healthy individuals performing four breathing exercises found that diaphragmatic breathing was the only one to significantly increase thoraco-abdominal asynchrony variables, simultaneously yielding the lowest thoracic contribution and highest abdominal contribution (PMID 25590447). The more “standard” the abdominal breathing, the less certain the instantaneous fluctuation in thoraco-abdominal coordination is. A design focused solely on $\phi$-SD will misread this normal fluctuation as “worsened coordination.”

    III. The Counter-Evidence Focuses on the “Form” of Variation, Not Its Size

    Conversely, there is evidence supporting the opposite view. The irregularity of conscious breathing correlates significantly with whether sleep apnea patients can adhere to CPAP therapy six months later (PMID 23080481)—irregularity itself carries information.

    The key to reconciling these lines is a 2023 study of 526 individuals across 11 tasks: the temporal structure of variability (lag-1 autocorrelation, power spectral density slope) shows high intrasubject retest consistency as a stable trait; however, it is nearly ineffective at explaining attentional states or self-rated traits; whereas more complex ARFIMA long-memory components show poor retest consistency (PMID 36618326). The “magnitude” of variability may not be the trait; the “form” of variability is more likely. If $\phi$-SD fails its own retest, the endpoint should shift to the autocorrelation structure of the $\phi$ sequence rather than stacking sample sizes on the SD.

    IV. Four Confounders That Must Be Excluded First

    1. Body Type. In supine position, abdominal volume proportion is 41% in obese groups versus 31% in normal-weight groups, while rib cage contribution is 5.7% versus 31.1% (PMID 37852999). Abdominal proportion directly determines the movable range of $\phi$.
    2. Posture. COPD patients exhibit thoracic asynchrony in sitting but not supine, whereas healthy controls show no partitioned asynchrony in either posture (PMID 23412901). A change in posture can flip sign-level conclusions for $\phi$.
    3. Breathing Frequency. Tidal volume low-frequency power share predicts up to 79% of HRV low-frequency power differences; authors argue respiratory variability must always be included in explanations (PMID 24423137). $\phi$-SD is highly collinear with respiratory cycle variability.
    4. Effective Sample Size. Respiratory sequences exhibit autocorrelation; independent sample counts are far fewer than breath counts, requiring effective degrees of freedom estimation for credible significance (PMID 39846001). An “SD” calculated from a 10-minute record has a confidence interval far wider than imagined.

    V. Design and Four-Grid Criteria

    $\ge$90 participants (novices / 1–5 years / >5 years, 30 each), fixed posture and inclination, continuous recording of $\ge$10 minutes of resting breathing. Extract $\phi$, thoracic and abdominal contributions, $V_T$, and respiratory cycle for each breath. Report circular dispersion using mean resultant length $R$ and circular variance, alongside arithmetic SD for comparison with prior literature. Match confounders by BMI, age, and gender; include respiratory cycle as a covariate.

    Four-grid judgment criteria:

    • Practitioners with small $\phi$-SD, and equally small $V_T$ and cycle variability, with $\phi$ mean centered away from $0^\circ$ → Overall Stability Established
    • Practitioners with small $\phi$-SD, but larger $V_T$ or cycle variability → Frozen Stability: sacrificing one dimension for overall stability is not deep skill
    • Practitioners with large $\phi$-SD, and the excess variability unrelated to ventilation efficiency → Functional Variation: the original hypothesis direction is incorrect; “liberation” does not equal degradation
    • No difference in $\phi$-SD between groups, but differences in lag-1 autocorrelation or spectral slope → Wrong Metric Selected; the endpoint must change; all prior $\phi$-SD inter-group comparisons downgrade to exploratory

    L1 Network Status: This topic has not yet entered the L1 corresponding network—CORRESPONDENCE_NETWORK.md contains no edges with “Dantian/Qigong/Thoraco-Abdominal Phase Angle/Respiratory Variability” as source or target nodes. Only physiological node tables list “Respiration Rate / Breathing Pattern” observations (annotated “most directly coupled with rhythm”) and “Gastrointestinal Peristalsis” as candidates for “Qi sinking to Dantian”; EDGE_TABLE_science-yixue.md only contains article counts (“Dantian 51 / Qigong 48 / Breathing 189”), lacking hierarchy (L1/L2/L3) and state (active/open/conflict) determinations. Per hard constraints, this article’s evidence level caps at C.

    Evidence and Limitations

    Verified: Thoraco-abdominal phase angle can be robustly measured via cross-correlation; thoraco-abdominal partition contributions show movement-state retest reliability ICC 0.87–0.93 (PMID 35226168); obese individuals show significantly higher abdominal proportion in supine position compared to normal-weight counterparts (PMID 37852999); posture changes can reverse thoracic asynchrony in COPD patients (PMID 23412901); diaphragmatic breathing significantly increases thoraco-abdominal asynchrony in healthy individuals (PMID 25590447); temporal structure of variability is more trait-like than magnitude (PMID 36618326).

    Partially Verified: Large intrasubject variability in $\phi$ stems from physiological instability (PMID 24716708, though sample was preterm infants); respiratory variability couples with HRV low-frequency power (PMID 24423137); the motor skill migration from freezing to release holds in movement tasks (PMID 32221040, 32670156), but has never been verified in respiratory tasks; conscious breathing irregularity correlates with clinical outcomes (PMID 23080481, single disease type).

    To Be Verified: The direction of difference in resting $\phi$-SD between practitioners and novices globally lacks direct research; adult retest reliability and minimal detectable change for $\phi$-SD; existence of “frozen stability” in the respiratory system; magnitude of bias between circular dispersion and arithmetic SD in crossing-zero scenarios.

    Fundamental Limitation: This article transplants variability theory from motor control to respiratory coordination, constituting analogical reasoning rather than direct evidence. Whether “coordination” in Dantian breathing shares the same degrees-of-freedom control mechanisms as motor skills remains untested. All conclusions remain design proposals prior to establishing $\phi$-SD retest reliability.

    Related Reading

    Evidence Level: C
    References: PMID 24716708, PMID 42862511, PMID 35226168, PMID 32221040, PMID 32670156, PMID 36618326, PMID 23080481, PMID 37852999, PMID 23412901, PMID 24423137, PMID 39846001, PMID 25590447, PMID 33848619


    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.