arxiv: 2605.06843 · v1 · submitted 2026-05-07 · 📊 stat.AP · stat.ME

Recognition: 2 theorem links

· Lean Theorem

Nonlinear Amplification of Finite-Sample Uncertainty in Capability-Based Decisions

Fei Jiang, Lei Yang

Pith reviewed 2026-05-11 01:10 UTC · model grok-4.3

classification 📊 stat.AP stat.ME

keywords process capability indicesfinite-sample uncertaintynonlinear amplificationdefect probabilityPPM ratesdecision reliabilitystatistical process control

0 comments

The pith

Finite-sample errors in capability indices amplify nonlinearly into defect probability estimates

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that uncertainty from finite samples in estimating process capability indices is transformed through a nonlinear mapping into defect probabilities and PPM rates. Small index variations that stay modest in linear scale become large swings in tail-risk measures because defect probability depends on distribution curvature. A reader cares because this explains why approval decisions based on capability indices can flip with modest sample changes, even when the indices themselves look stable. The work ties sample-size needs directly to decision reliability and validates the effect with simulations and industrial data.

Core claim

Capability estimators vary approximately linearly with process dispersion, but defect probabilities depend on tail curvature, causing small estimation errors to be disproportionately amplified in defect probability and parts-per-million rates, so that capability assessments that appear stable in index space exhibit substantial variability in defect-risk space particularly near decision thresholds.

What carries the argument

The nonlinear amplification mechanism that maps linear capability-index uncertainty through tail curvature into defect-probability variability.

If this is right

Sample-size planning must be linked to required decision reliability in defect-risk space rather than index precision alone.
Capability-based approvals need explicit uncertainty propagation to defect metrics to avoid threshold instability.
Reliability-aware decision rules are required to account for the amplified variability observed in simulations and real data.
The mechanism supplies a unified account of finite-sample instabilities that appear in manufacturing practice.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same tail-curvature amplification may appear in other statistical decisions that convert linear estimates into probabilities, such as reliability or risk thresholds.
Moving decision boundaries away from steep regions of the defect-probability curve could reduce observed instability.
The framework invites explicit sample-size formulas that target acceptable variance in PPM rates directly.

Load-bearing premise

The dominant source of decision instability is the nonlinear mapping from capability index to tail probability rather than model misspecification or non-stationarity in the process data.

What would settle it

Monte Carlo runs or industrial datasets in which defect-probability variability matches only the linear propagation of index variance, with no extra spread from tail curvature near thresholds, would falsify the claimed amplification.

Figures

Figures reproduced from arXiv: 2605.06843 by Fei Jiang, Lei Yang.

**Figure 1.** Figure 1: shows how acceptance probability evolves with sample size under different values of C true pk . When 10 20 30 40 50 60 70 80 Sample Size (n) 0.0 0.2 0.4 0.6 0.8 1.0 p a c c(n) nmin(0:95; 1:67) = 26 C true pk 1.00 1.33 1.67 [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗

**Figure 2.** Figure 2: Model-induced capability distortion ∆Cpk = Cbpk − CbNormal pk across the 18 selected dimensions. Negative values indicate inflation of capability under the normal assumption relative to the fitted distribution; positive values indicate conservative bias. The observed distortions, while moderate in capability space, can lead to amplified differences in defect-risk metrics and decision outcomes. Although Eq… view at source ↗

read the original abstract

This paper studies the propagation of finite-sample uncertainty under nonlinear transformations commonly used in statistical decision systems. In particular, we consider process capability indices, which are widely used in manufacturing practice but are estimated from finite samples, rendering the resulting approval decisions inherently uncertain. We show that such uncertainty cannot be fully explained by estimator variability alone, but is substantially influenced by a nonlinear amplification mechanism through which capability uncertainty is transformed into defect-risk metrics. While capability estimators vary approximately linearly with process dispersion, defect probabilities depend on tail curvature, causing small estimation errors to be disproportionately amplified in measures such as defect probability and parts-per-million (PPM) rates. Consequently, capability assessments that appear stable in index space may exhibit substantial variability in defect-risk space, particularly near decision thresholds. This insight provides a unified explanation of finite-sample decision instability, motivates reliability-aware decision formulations, and links sample-size requirements directly to decision reliability. Monte Carlo simulations and industrial data analyses validate the proposed mechanism and demonstrate its practical implications, including the impact of distributional assumptions on defect-risk estimation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper shows how nonlinearity in defect-probability tails amplifies small capability-index errors into large decision instability, but the Monte Carlo and data checks do not clearly separate this from plain estimator variance.

read the letter

The one thing to know is that the authors argue small estimation errors in Cp or Cpk get blown up when turned into PPM or defect rates because the tail mapping is nonlinear, and they claim this explains instability near approval thresholds better than variance alone. They tie it to sample-size planning for reliable binary decisions in manufacturing. That linkage is the concrete contribution. The paper does a clean job laying out the basic mechanism: capability estimators move roughly linearly with dispersion, but the probability transform curves sharply, so modest index jitter produces outsized risk jitter. The Monte Carlo runs and industrial examples are meant to illustrate the practical size of the effect. That framing is useful for people who already use capability indices and want a reason to collect larger samples near decision boundaries. The soft spot is the attribution. The abstract says the instability cannot be fully explained by estimator variability alone and is substantially influenced by the nonlinear step, yet the description gives no quantitative comparison to a linear approximation or delta-method baseline on the same simulated or real data. Without that isolation, it is hard to tell how much of the observed PPM spread comes from the claimed curvature versus ordinary sampling noise or unmodeled process shifts. If the full text contains those side-by-side numbers and error bars, the claim strengthens; otherwise the central evidence remains suggestive rather than decisive. This is for applied statisticians and quality engineers who make go/no-go calls with capability data. A reader who needs to justify larger sample sizes or add uncertainty buffers will get a usable story. It is coherent enough and addresses a real industrial pain point, so it deserves a serious referee who can press on the simulation design and ask for the missing linear baseline.

Referee Report

2 major / 2 minor

Summary. The paper claims that finite-sample uncertainty in process capability index estimators propagates through nonlinear tail mappings (e.g., to defect probabilities and PPM rates) in a manner that cannot be fully explained by estimator variability alone. Small errors in the capability index are disproportionately amplified near decision thresholds due to tail curvature, producing substantial instability in defect-risk metrics even when index values appear stable. Monte Carlo simulations and industrial data analyses are presented as validation, with implications for reliability-aware decisions and sample-size requirements tied to decision reliability.

Significance. If the nonlinear amplification mechanism can be isolated from other sources of variability, the work supplies a coherent explanation for observed instabilities in capability-based manufacturing decisions and directly connects sample-size planning to decision reliability. The attempt to unify estimator variability with tail nonlinearity and to demonstrate practical consequences via both simulation and real data is a constructive contribution to applied statistics in quality control.

major comments (2)

[Monte Carlo Simulations] Monte Carlo Simulations section: the central claim that uncertainty 'cannot be fully explained by estimator variability alone' and is 'substantially influenced' by nonlinear amplification requires an explicit baseline comparison (e.g., delta-method or first-order Taylor expansion of the tail probability applied to the same estimated indices) to quantify the incremental contribution of tail curvature versus inherent estimator variance. No such comparison is described, leaving the dominance of the nonlinear mechanism unverified.
[Industrial Data Analyses] Industrial Data Analyses section: the validation asserts support from real data but supplies no quantitative results, error bars, confidence intervals, or description of controls for confounding factors such as non-stationarity or model misspecification. Without these, the degree of empirical support for the amplification claim cannot be assessed.

minor comments (2)

[Abstract] Abstract: the summary of validation results is entirely qualitative; adding one or two key quantitative findings (e.g., observed amplification factors or PPM variability ranges) would improve informativeness.
[Introduction] Notation: the distinction between linear variation in the capability estimator and nonlinear tail mapping is stated clearly in prose but would benefit from an explicit equation contrasting the two mappings (e.g., a first-order vs. full nonlinear expression for defect probability).

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the insightful comments on our manuscript. We address each of the major comments below, indicating the changes we plan to make in the revised version.

read point-by-point responses

Referee: [Monte Carlo Simulations] Monte Carlo Simulations section: the central claim that uncertainty 'cannot be fully explained by estimator variability alone' and is 'substantially influenced' by nonlinear amplification requires an explicit baseline comparison (e.g., delta-method or first-order Taylor expansion of the tail probability applied to the same estimated indices) to quantify the incremental contribution of tail curvature versus inherent estimator variance. No such comparison is described, leaving the dominance of the nonlinear mechanism unverified.

Authors: We agree that providing an explicit baseline comparison would better isolate the contribution of the nonlinear amplification mechanism. In the revised manuscript, we will augment the Monte Carlo Simulations section with a comparison against a first-order approximation (such as the delta method applied to the tail probability function) using the same estimated capability indices. This will allow us to quantify the additional variability introduced by the tail curvature beyond what is attributable to estimator variance alone. We believe this addition will substantiate our central claim more rigorously. revision: yes
Referee: [Industrial Data Analyses] Industrial Data Analyses section: the validation asserts support from real data but supplies no quantitative results, error bars, confidence intervals, or description of controls for confounding factors such as non-stationarity or model misspecification. Without these, the degree of empirical support for the amplification claim cannot be assessed.

Authors: We acknowledge the need for more detailed quantitative reporting in the Industrial Data Analyses section. In the revision, we will include specific numerical results from the data analyses, along with error bars and confidence intervals for the key defect-risk metrics. Additionally, we will describe the steps taken to address potential confounding factors, including checks for stationarity and sensitivity to distributional assumptions. This will enable a better evaluation of the empirical validation provided by the industrial data. revision: yes

Circularity Check

0 steps flagged

No circularity: mathematical distinction and simulation validation are self-contained

full rationale

The paper's core argument distinguishes linear finite-sample variability of capability-index estimators from the nonlinear curvature of the tail-probability mapping (e.g., defect probability or PPM). This distinction is presented as a direct mathematical property of the transformation, not derived from any fitted parameter or self-referential definition. Validation is supplied by Monte Carlo experiments and industrial data that compare observed variability against the claimed mechanism; no equation is shown to reduce the target quantity to a quantity defined by the same data or by a prior self-citation. No self-citation load-bearing steps, uniqueness theorems, or ansatz smuggling are indicated in the provided text. The derivation therefore remains independent of its inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard statistical assumptions about sampling variability and the use of tail probabilities to compute defect rates; no new free parameters or invented entities are introduced in the abstract.

axioms (2)

domain assumption Capability indices are estimated from finite samples drawn from a process whose distribution permits tail-probability calculations.
This is the source of the finite-sample uncertainty that is then transformed nonlinearly.
domain assumption Defect probability is obtained by evaluating the tail of the fitted distribution beyond specification limits.
This step supplies the curvature responsible for amplification.

pith-pipeline@v0.9.0 · 5471 in / 1363 out tokens · 31342 ms · 2026-05-11T01:10:55.716353+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
Aσ = z ϕ(z) / (1−Φ(z)) = z r(z) ... Aσ = 3 Cpk r(3 Cpk)
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear
Var([PPM) ≈ ∇h⊤ Σ ∇h ... Var([PPM) ≫ Var(Ĉpk)

Reference graph

Works this paper leans on

32 extracted references · 32 canonical work pages · 1 internal anchor

[1]

Victor E. Kane. Process Capability Indices.Jour- nal of Quality Technology, 18(1):41–52, January

work page
[2]

doi: 10.1080/00224065.1986

ISSN 0022-4065. doi: 10.1080/00224065.1986. 11978984

work page doi:10.1080/00224065.1986 1986
[3]

Mohammed Z. Anis. Basic Process Capability In- dices: An Expository Review.International Statis- tical Review, 76(3):347–367, December 2008. ISSN 0306-7734, 1751-5823. doi: 10.1111/j.1751-5823. 2008.00060.x

work page doi:10.1111/j.1751-5823 2008
[4]

Samuel Kotz and Norman L. Johnson. Process Ca- pability Indices—A Review, 1992–2000.Journal of Quality Technology, 34(1):2–19, January 2002. ISSN 0022-4065, 2575-6230. doi: 10.1080/00224065.2002. 11980119

work page doi:10.1080/00224065.2002 1992
[5]

John wiley & sons, 2020

Douglas C Montgomery.Introduction to statistical quality control. John wiley & sons, 2020

work page 2020
[6]

An overview of theory and practice on process capability indices for quality assurance.International journal of production economics, 117(2):338–359, 2009

Chien-Wei Wu, WL Pearn, and Samuel Kotz. An overview of theory and practice on process capability indices for quality assurance.International journal of production economics, 117(2):338–359, 2009

work page 2009
[7]

Statisticalmethodsinprocessmanagement – capability and performance – part 4: Process capa- bility estimates and performance measures

ISO/TR. Statisticalmethodsinprocessmanagement – capability and performance – part 4: Process capa- bility estimates and performance measures. ISO/TR 22514-4:2016 (2016)

work page 2016
[8]

Guidelines for implementation of statistical process control (spc) - part 1: Elements of spc

ISO. Guidelines for implementation of statistical process control (spc) - part 1: Elements of spc. ISO 11462-1:2001 (2001)

work page 2001
[9]

Statisticalmethodsinprocessmanagement 9 – capability and performance – part 1: General prin- ciples and concepts

ISO/TR. Statisticalmethodsinprocessmanagement 9 – capability and performance – part 1: General prin- ciples and concepts. ISO/TR 22514-1:2014 (2014)

work page 2014
[10]

W. L. Pearn, Samuel Kotz, and Norman L. Johnson. Distributional and Inferential Properties of Process Capability Indices.Journal of Quality Technology, 24(4):216–231, October 1992. ISSN 0022-4065, 2575-

work page 1992
[11]

doi: 10.1080/00224065.1992.11979403

work page doi:10.1080/00224065.1992.11979403 1992
[12]

Confidence bounds for capability indices.Journal of Quality Technology, 24(4):188–195, 1992

Robert H Kushler and Paul Hurley. Confidence bounds for capability indices.Journal of Quality Technology, 24(4):188–195, 1992

work page 1992
[13]

Generalized confidence intervals for process capability indices.Quality and reliability engineering international, 23(4):471–481, 2007

Thomas Mathew, George Sebastian, and KM Kurian. Generalized confidence intervals for process capability indices.Quality and reliability engineering international, 23(4):471–481, 2007

work page 2007
[14]

Interval estimation of process capability index cpk.Commu- nications in Statistics-Theory and Methods, 19(12): 4455–4470, 1990

NF Zhang, GA Stenback, and DM Wardrop. Interval estimation of process capability index cpk.Commu- nications in Statistics-Theory and Methods, 19(12): 4455–4470, 1990

work page 1990
[15]

Mahmoud, G

Mahmoud A. Mahmoud, G. Robin Henderson, Eu- genio K. Epprecht, and William H. Woodall. Esti- mating the Standard Deviation in Quality-Control Applications.Journal of Quality Technology, 42(4): 348–357, October 2010. ISSN 0022-4065, 2575-6230. doi: 10.1080/00224065.2010.11917832

work page doi:10.1080/00224065.2010.11917832 2010
[16]

Moya-Férnandez, Francisco J

Encarnación Álvarez, Pablo J. Moya-Férnandez, Francisco J. Blanco-Encomienda, and Juan F. Muñoz. Methodological insights for industrial qual- ity control management: The impact of various es- timators of the standard deviation on the process capability index.Journal of King Saud University - Science, 27(3):271–277, July 2015. ISSN 10183647. doi: 10.1016/j...

work page doi:10.1016/j.jksus.2015.02.002 2015
[17]

K. S. Chen and W. L. Pearn. An ap- plication of non-normal process capability in- dices.Quality and Reliability Engineering In- ternational, 13(6):355–360, 1997. ISSN 1099-

work page 1997
[18]

doi: 10.1002/(SICI)1099-1638(199711/12)13: 6<355::AID-QRE125>3.0.CO;2-V

work page doi:10.1002/(sici)1099-1638(199711/12)13:
[19]

Jann-Pygn Chen and Cherng G. Ding. A new pro- cess capability index for non-normal distributions. International Journal of Quality & Reliability Man- agement, 18(7):762–770, October 2001. ISSN 0265- 671X. doi: 10.1108/02656710110396076

work page doi:10.1108/02656710110396076 2001
[20]

Process capability calculations for non-normal distributions.Quality progress, 22: 95–100, 1989

John A Clements. Process capability calculations for non-normal distributions.Quality progress, 22: 95–100, 1989

work page 1989
[21]

Bootstrap confidence limits on pro- cess capability indices.Journal of the Royal Sta- tistical Society: Series D (The Statistician), 44(3): 373–378, 1995

Alan J Collins. Bootstrap confidence limits on pro- cess capability indices.Journal of the Royal Sta- tistical Society: Series D (The Statistician), 44(3): 373–378, 1995

work page 1995
[22]

Finite-sample decision insta- bility in threshold-based process capability approval

Fei Jiang and Lei Yang. Finite-sample decision insta- bility in threshold-based process capability approval. arXiv:2603.11315, 2026

work page arXiv 2026
[23]

How reliable is your capability index? Journal of the Royal Statistical Society Series C: Ap- plied Statistics, 39(3):331–340, 1990

AF Bissell. How reliable is your capability index? Journal of the Royal Statistical Society Series C: Ap- plied Statistics, 39(3):331–340, 1990

work page 1990
[24]

Practical process capa- bility indices workflows.The International Jour- nal of Advanced Manufacturing Technology, pages 1–19, 2026

Fei Jiang and Lei Yang. Practical process capa- bility indices workflows.The International Jour- nal of Advanced Manufacturing Technology, pages 1–19, 2026. doi: 10.1007/s00170-026-17782-7. URL https://doi.org/10.1007/s00170-026-17782-7

work page doi:10.1007/s00170-026-17782-7 2026
[25]

Risk-calibrated process ca- pability approval with finite samples.arXiv preprint arXiv:2603.14479, 2026

Fei Jiang and Lei Yang. Risk-calibrated process ca- pability approval with finite samples.arXiv preprint arXiv:2603.14479, 2026

work page internal anchor Pith review arXiv 2026
[26]

Using measurement uncer- tainty in decision-making and conformity assess- ment.Metrologia, 51(4):S206–S218, 2014

Leslie R Pendrill. Using measurement uncer- tainty in decision-making and conformity assess- ment.Metrologia, 51(4):S206–S218, 2014

work page 2014
[27]

Use of the ppm and its function in the production process.Procedia Manufacturing, 13:608–615, 2017

L Bebr, K Bícová, and H Zídková. Use of the ppm and its function in the production process.Procedia Manufacturing, 13:608–615, 2017

work page 2017
[28]

On the gap between theory and prac- tice of process capability studies.International Jour- nal of Quality & Reliability Management, 15(2):178– 191, 1998

Mats Deleryd. On the gap between theory and prac- tice of process capability studies.International Jour- nal of Quality & Reliability Management, 15(2):178– 191, 1998

work page 1998
[29]

John Wiley & Sons, 2005

Morris H DeGroot.Optimal statistical decisions. John Wiley & Sons, 2005

work page 2005
[30]

Springer, 2005

Erich Leo Lehmann and Joseph P Romano.Testing statistical hypotheses. Springer, 2005

work page 2005
[31]

Springer Science & Business Me- dia, 2013

James O Berger.Statistical decision theory and Bayesian analysis. Springer Science & Business Me- dia, 2013

work page 2013
[32]

Cambridge university press, 2000

Aad W Van der Vaart.Asymptotic statistics, vol- ume 3. Cambridge university press, 2000. 10

work page 2000