pith. sign in

arxiv: 2605.20135 · v1 · pith:EM2E4L35new · submitted 2026-05-19 · 📊 stat.ME

Quantile-Based Effectiveness Persistence Function: A Tail-Focused Metric with Theory, Estimation, and Application to Biosimilar Evaluation

Sankaran P.G. , Prasanth V.P. , Midhu N.N This is my paper

Pith reviewed 2026-05-20 03:20 UTC · model grok-4.3

classification 📊 stat.ME
keywords quantile-based persistencetail meanupper tail analysisL-momentbiosimilar evaluationnonparametric estimatorequivalence testclinical adherence
0
0 comments X

The pith

A ratio of tail mean to quantile function measures how long high clinical responses persist among top performers.

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

The paper defines a quantile-based effectiveness persistence function as the ratio of the tail mean to the quantile function. This construction parallels expected shortfall and equals the first L-moment of the scaled tail. The authors prove its properties, supply a nonparametric estimator, and build a bootstrap-calibrated two-sample equivalence test. The metric is intended to detect clinically relevant deviations in the upper tail of response distributions that conventional mean or median summaries miss. It is applied to biosimilar evaluation where sustained high performance matters more than average behavior.

Core claim

The quantile-based effectiveness persistence function is defined as the ratio between the tail mean and the quantile function. It is shown to be equivalent to the first L-moment of the scaled tail and therefore supplies robust inference tools for upper-tail analysis. A simple nonparametric estimator is derived and paired with a bootstrap-calibrated two-sample equivalence test that targets clinically meaningful tail differences.

What carries the argument

The quantile-based effectiveness persistence function, defined as the ratio of tail mean to quantile function, quantifies persistence in the upper tail of clinical response distributions.

If this is right

  • The function complements median and mean summaries by focusing on how long high responses persist.
  • It yields a bootstrap-based two-sample test for equivalence in upper-tail behavior.
  • Nonparametric estimation allows application to biosimilar data without parametric tail assumptions.
  • Properties derived from its equivalence to the first L-moment support stable inference in small samples.

Where Pith is reading between the lines

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

  • The same tail-ratio idea could be adapted to other fields that track sustained extreme performance, such as reliability engineering.
  • Regulatory comparisons of biosimilars might incorporate this metric to emphasize long-term top-end outcomes rather than averages.
  • Repeated application across multiple quantiles could reveal whether persistence is uniform or concentrated at particular tail levels.

Load-bearing premise

The upper tail of the response distribution in clinical data captures clinically relevant persistence that this ratio can measure without further distributional assumptions.

What would settle it

In a controlled simulation or real dataset with known upper-tail differences, the new ratio fails to detect the difference while standard mean or median tests succeed, or vice versa.

read the original abstract

In clinical studies, persistence, which measures the duration of time a patient continues to take a prescribed medication without discontinuation, is increasingly recognized as a critical indicator of adherence to medication. Adherence encompasses not only whether a patient takes their medication as prescribed but also the consistency and duration with which they do so. Among the various metrics used to evaluate adherence, persistence stands out as a particularly robust measure because it provides a temporal dimension, reflecting the sustained commitment of patients to their therapeutic regimens. This focus on persistence offers unique insights into adherence-related quality and performance, shedding light on the challenges and opportunities to optimize long-term medication use. The comparison of upper-tail clinical performance, which measures the extent to which very large responses persist among top responders, is often more decisive in therapy evaluation than conventional summaries. In this paper, we introduce the quantile-based effectiveness persistence function defined as the ratio between the tail mean and the quantile function. The notion parallels expected shortfall in risk theory and is tailored to detect clinically meaningful deviations in the upper tail. We establish key properties and show that the function is equivalent to the first L-moment of the scaled tail, yielding robust inference tools. We derive a simple nonparametric estimator of the function and develop a bootstrap-calibrated two-sample (upper-tail) equivalence test. Simulation studies and real-data analysis illustrate that the proposed measures captures clinically relevant tail persistence that complements median and mean-based summaries.

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.

Referee Report

2 major / 1 minor

Summary. The paper introduces the quantile-based effectiveness persistence function, defined as the ratio of the tail mean to the quantile function Q(p). It claims this metric parallels expected shortfall, is equivalent to the first L-moment of the scaled tail, and provides a nonparametric estimator along with a bootstrap-calibrated two-sample equivalence test for upper-tail deviations. The approach is motivated by clinical persistence in medication adherence and illustrated via simulations and biosimilar data analysis.

Significance. If the estimator's properties hold in moderate samples, the metric could offer a useful complement to mean- and median-based summaries for detecting clinically relevant tail persistence in clinical trials, particularly for biosimilar evaluations where upper-tail performance matters.

major comments (2)
  1. [Theory section] The equivalence to the first L-moment of the scaled tail is a direct definitional identity once the scaled tail is taken as X/Q(p) conditional on X > Q(p); this should be stated explicitly in the theory section rather than presented as an independent result.
  2. [Estimation and simulation studies] The nonparametric estimator of the tail-mean/quantile ratio inherits the slow convergence of extreme quantiles. In biosimilar trials with n typically 100-300 per arm and p near 1, this risks unstable estimates and degraded bootstrap coverage under heavy tails, which directly affects the reliability of the proposed two-sample equivalence test.
minor comments (1)
  1. Clarify the choice of p in applications and provide guidance on sensitivity to this tuning parameter.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment below and indicate the revisions we will make to strengthen the presentation and address practical concerns.

read point-by-point responses

  1. Referee: [Theory section] The equivalence to the first L-moment of the scaled tail is a direct definitional identity once the scaled tail is taken as X/Q(p) conditional on X > Q(p); this should be stated explicitly in the theory section rather than presented as an independent result.

    Authors: We agree that the stated equivalence follows directly once the scaled tail is defined as the conditional random variable X/Q(p) given X > Q(p). In the revised manuscript we will explicitly note this as a definitional identity in the theory section and adjust the surrounding text to avoid any implication that it is an independent result. This change improves clarity without altering the technical content. revision: yes

  2. Referee: [Estimation and simulation studies] The nonparametric estimator of the tail-mean/quantile ratio inherits the slow convergence of extreme quantiles. In biosimilar trials with n typically 100-300 per arm and p near 1, this risks unstable estimates and degraded bootstrap coverage under heavy tails, which directly affects the reliability of the proposed two-sample equivalence test.

    Authors: The referee correctly notes that the estimator inherits the slower convergence rate of extreme-quantile estimators. Our existing simulation design already includes sample sizes in the 100–300 range and several tail behaviors. To respond directly to the concern, we will add a new subsection that reports bootstrap coverage under heavier-tailed distributions (e.g., Student-t with low degrees of freedom) and will include a brief discussion of practical guidance for selecting p when n is moderate. These additions will be presented as supplementary simulation results. revision: partial

Circularity Check

1 steps flagged

Equivalence of persistence function to first L-moment is definitional identity

specific steps
  1. self definitional [Abstract]
    "we introduce the quantile-based effectiveness persistence function defined as the ratio between the tail mean and the quantile function. ... We establish key properties and show that the function is equivalent to the first L-moment of the scaled tail, yielding robust inference tools."

    The function is defined as tail mean / quantile. The scaled tail is X/Q(p) conditional on X > Q(p), so its mean equals the defined ratio. The first L-moment of any distribution is its mean, making the stated equivalence true by construction from the definition plus the known fact that λ1 = mean. This is not an independent result but a direct restatement.

full rationale

The paper defines the quantile-based effectiveness persistence function explicitly as the ratio of tail mean to quantile function. It then presents the equivalence to the first L-moment of the scaled tail as a key property that yields robust inference. Because the first L-moment equals the mean and the ratio is precisely the mean of the scaled tail (X/Q(p) | X > Q(p)), the equivalence holds by algebraic identity from the definition and standard L-moment properties rather than an independent derivation. This makes the central claim of 'yielding robust inference tools' reduce to a renaming of the input definition. The nonparametric estimator and bootstrap test appear to have separate content, so the circularity is partial rather than total. No load-bearing self-citations or other patterns are evident in the provided text.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The claim relies on standard properties of quantiles and L-moments from statistical theory, with the new function defined directly.

axioms (1)
  • domain assumption The tail mean and quantile function are well-defined for the positive response distributions in clinical data.
    Required for the ratio to be meaningful.
invented entities (1)
  • quantile-based effectiveness persistence function no independent evidence
    purpose: Measure of upper-tail persistence in effectiveness
    Introduced in this paper as a new metric.

pith-pipeline@v0.9.0 · 5793 in / 1315 out tokens · 58794 ms · 2026-05-20T03:20:54.420743+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

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

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

134 extracted references · 134 canonical work pages · 1 internal anchor

  1. [1]

    and Moeschberger, Melvin L

    Klein, John P. and Moeschberger, Melvin L. , title =. 2003 , edition =

    work page 2003

  2. [2]

    , title =

    Arnold, Barry C. , title =. 2015 , publisher =

    work page 2015

  3. [3]

    Unnikrishnan and Sankaran, P

    Nair, N. Unnikrishnan and Sankaran, P. G. and Balakrishnan, N. , title =. doi:10.1007/978-0-8176-8361-0 , publisher =

    work page doi:10.1007/978-0-8176-8361-0

  4. [4]

    Mathematical Finance , volume =

    Artzner, Philippe and Delbaen, Freddy and Eber, Jean-Marc and Heath, David , title =. Mathematical Finance , volume =

    work page

  5. [5]

    Bahadur, R. R. , title =. Annals of Mathematical Statistics , volume =

    work page

  6. [6]

    and Proschan, Frank , title =

    Barlow, Richard E. and Proschan, Frank , title =

    work page

  7. [7]

    The total time on test concept and its use in reliability theory , journal =

    Bergman, Bo and Klefsj. The total time on test concept and its use in reliability theory , journal =

    work page

  8. [8]

    Annals of Statistics , volume =

    Chernozhukov, Victor and Chetverikov, Denis and Kato, Kengo , title =. Annals of Statistics , volume =

    work page

  9. [9]

    Davison, A. C. and Hinkley, D. V. , title =

    work page

  10. [10]

    de Haan, Laurens and Ferreira, Ana , title =

    work page

  11. [11]

    Efron, Bradley and Tibshirani, Robert , title =

    work page

  12. [12]

    Guideline on similar biological medicinal products , howpublished =

    work page

  13. [13]

    1997 , publisher =

    Modelling extremal events for insurance and finance , author =. 1997 , publisher =

    work page 1997

  14. [14]

    Scientific considerations in demonstrating biosimilarity to a reference product , howpublished =

    work page

  15. [15]

    2011 , note =

    Foss, Serguei and Korshunov, Dmitry and Zachary, Stan , title =. 2011 , note =

    work page 2011

  16. [16]

    1985 , url =

    Guess, Frank and Proschan, Frank , title =. 1985 , url =

    work page 1985

  17. [17]

    and Davis, Charles E

    Harrell, Frank E. and Davis, Charles E. , title =. Biometrika , volume =

    work page

  18. [18]

    Jeong, Jong-Hyeon , title =

    work page

  19. [19]

    Kim, K. H. and Kang, S. and Chiou, S. H. , title =. The R Journal , volume =

    work page

  20. [20]

    Annals of Mathematical Statistics , volume =

    Kiefer, Jack , title =. Annals of Mathematical Statistics , volume =

    work page

  21. [21]

    Journal of Statistical Software , volume =

    Koenker, Roger , title =. Journal of Statistical Software , volume =

    work page

  22. [22]

    Unnikrishnan and Sankaran, P

    Nair, N. Unnikrishnan and Sankaran, P. G. and Balakrishnan, N. , title =

    work page

  23. [23]

    and Staudte, Robert G

    Prendergast, Luke A. and Staudte, Robert G. , title =. Electronic Journal of Statistics , volume =

    work page

  24. [24]

    and Rotnitzky, Andrea , title =

    Robins, James M. and Rotnitzky, Andrea , title =. AIDS Epidemiology , pages =

    work page

  25. [25]

    , title =

    Serfling, Robert J. , title =

    work page

  26. [26]

    van der Vaart, A. W. , title =

    work page

  27. [27]

    van der Vaart, A. W. and Wellner, J. A. , title =

    work page

  28. [28]

    2014 , url =

    Isogai, Takashi , title =. 2014 , url =

    work page 2014

  29. [29]

    and Frey, R

    McNeil, Alexander J. and Frey, R. Quantitative risk management: Concepts, techniques, and tools , publisher =

    work page

  30. [30]

    Annals of Probability , volume =

    Residual life time at great age , author =. Annals of Probability , volume =. 1974 , publisher =

    work page 1974

  31. [31]

    Journal of Statistical Planning and Inference , volume =

    Fixed values versus empirical quantiles as thresholds in excess distribution modelling , author =. Journal of Statistical Planning and Inference , volume =. 2025 , publisher =

    work page 2025

  32. [32]

    2003 , edition =

    Order statistics , author =. 2003 , edition =

    work page 2003

  33. [33]

    2001 , publisher =

    An introduction to statistical modeling of extreme values , author =. 2001 , publisher =

    work page 2001

  34. [34]

    Statistical Methodology , volume =

    A class of distributions with the linear mean residual quantile function and its generalizations , author =. Statistical Methodology , volume =. 2013 , publisher =

    work page 2013

  35. [35]

    e-Handbook of Statistical Methods: Beta distribution (general form with bounds a,b ) , year =

    work page

  36. [36]

    and Nadarajah, S

    Aryal, G. and Nadarajah, S. , title =. Serdica Mathematical Journal , volume =. 2004 , note =

    work page 2004

  37. [37]

    Advances in Data Analysis and Classification , year =

    Bee, Marco , title =. Advances in Data Analysis and Classification , year =. doi:10.1007/s11634-022-00497-4 , url =

    work page doi:10.1007/s11634-022-00497-4

  38. [38]

    Journal of Statistical Theory and Practice , year =

    Ando, Tomohiro , title =. Journal of Statistical Theory and Practice , year =. doi:10.1007/s42519-022-00202-3 , url =

    work page doi:10.1007/s42519-022-00202-3

  39. [39]

    Cover and Joy A

    Cover, Thomas M. and Thomas, Joy A. , title =. 2006 , isbn =. doi:10.1002/047174882X , url =

    work page doi:10.1002/047174882x 2006

  40. [40]

    , title =

    Baraty, Saaid and Simovici, Dan A. , title =. Proceedings of the 23rd International Florida Artificial Intelligence Research Society Conference (FLAIRS-23) , pages =. 2010 , publisher =

    work page 2010

  41. [41]

    Bernoulli , year =

    Gulisashvili, Archil and Tankov, Peter , title =. Bernoulli , year =. doi:10.3150/14-BEJ665 , url =

    work page doi:10.3150/14-bej665

  42. [42]

    Applied probability and queues , edition =

    Asmussen, S. Applied probability and queues , edition =. 2003 , series =. doi:10.1007/b97236 , url =

    work page doi:10.1007/b97236 2003

  43. [43]

    2022 , series =

    Nair, Jayakrishnan and Wierman, Adam and Zwart, Bert , title =. 2022 , series =. doi:10.1017/9781009053730 , url =

    work page doi:10.1017/9781009053730 2022

  44. [44]

    Mean residual life processes , journal =

    Cs. Mean residual life processes , journal =. 1996 , volume =

    work page 1996

  45. [45]

    Everest, Louis and Shah, Monica and Chan, Kelvin K. W. , title =. JAMA Network Open , year =. doi:10.1001/jamanetworkopen.2019.6803 , url =

    work page doi:10.1001/jamanetworkopen.2019.6803 2019

  46. [46]

    1998 , url =

    E9: Statistical principles for clinical trials , howpublished =. 1998 , url =

    work page 1998

  47. [47]

    Sankaran, P. G. and Midhu, N. N. , title =. Journal of Applied Statistics , volume =. 2017 , doi =

    work page 2017

  48. [48]

    Midhu, N. N. and Sankaran, P. G. and Nair, N. Unnikrishnan , title =. Statistical Methodology , year =. doi:10.1016/j.stamet.2013.03.002 , note =

    work page doi:10.1016/j.stamet.2013.03.002 2013

  49. [49]

    Journal of Banking & Finance , year =

    Acerbi, Carlo and Tasche, Dirk , title =. Journal of Banking & Finance , year =

    work page

  50. [50]

    Aarset, M. V. , title =. IEEE Transactions on Reliability , year =

    work page

  51. [51]

    Bingham, N. H. and Goldie, C. M. and Teugels, J. L. , title =. 1987 , publisher =

    work page 1987

  52. [52]

    The Annals of Statistics , year =

    Chernozhukov, Victor and Chetverikov, Denis and Kato, Kengo , title =. The Annals of Statistics , year =. doi:10.1214/13-AOS1161 , url =

    work page doi:10.1214/13-aos1161

  53. [53]

    2013 , publisher =

    Foss, Sergey and Korshunov, Dmitry and Zachary, Stanislav , title =. 2013 , publisher =

    work page 2013

  54. [54]

    , title =

    Gastwirth, Joseph L. , title =. Econometrica , year =. doi:10.2307/1909675 , url =

    work page doi:10.2307/1909675

  55. [55]

    Gastwirth , title =

    Joseph L. Gastwirth , title =. The Review of Economics and Statistics , volume =. 1972 , doi =

    work page 1972

  56. [56]

    Hosking, J. R. M. , title =. Journal of the Royal Statistical Society. Series B (Methodological) , year =

    work page

  57. [57]

    E9(R1) Addendum on Estimands and Sensitivity Analysis in Clinical Trials to the Guideline on Statistical Principles for Clinical Trials , year =

    work page

  58. [58]

    and Prentice, Ross L

    Kalbfleisch, John D. and Prentice, Ross L. , title =. 2002 , publisher =

    work page 2002

  59. [59]

    Regression quantiles,

    Koenker, Roger and Bassett, Gilbert , title =. Econometrica , year =. doi:10.2307/1913643 , url =

    work page doi:10.2307/1913643

  60. [60]

    2005 , publisher =

    Koenker, Roger , title =. 2005 , publisher =

    work page 2005

  61. [61]

    and Frey, R

    McNeil, Alexander J. and Frey, R. Quantitative Risk Management: Concepts, Techniques and Tools , edition =. 2015 , publisher =

    work page 2015

  62. [62]

    Tyrrell and Uryasev, Stanislav , title =

    Rockafellar, R. Tyrrell and Uryasev, Stanislav , title =. Journal of Banking & Finance , year =. doi:10.1016/S0378-4266(02)00271-6 , url =

    work page doi:10.1016/s0378-4266(02)00271-6

  63. [63]

    and Grambsch, Patricia M

    Therneau, Terry M. and Grambsch, Patricia M. and others , title =. 2025 , organization =

    work page 2025

  64. [64]

    Non-Inferiority Clinical Trials to Establish Effectiveness: Guidance for Industry , year =

    work page

  65. [65]

    Kalbfleisch, J. D. and Prentice, R. L. , title =. 2002 , isbn =

    work page 2002

  66. [66]

    Lawless , title =

    Jerald F. Lawless , title =. 2003 , publisher =

    work page 2003

  67. [67]

    Johnson and Samuel Kotz and N

    Norman L. Johnson and Samuel Kotz and N. Balakrishnan , title =. 1994 , publisher =

    work page 1994

  68. [68]

    Reliability Engineering & System Safety , year =

    Saralees Nadarajah and Samuel Kotz , title =. Reliability Engineering & System Safety , year =

    work page

  69. [69]

    Fleming, T. R. and Harrington, D. P. , title =

    work page

  70. [70]

    Gill, R. D. , title =. Ann. Statist. , year =

    work page

  71. [71]

    2025 , note =

    German Breast Cancer Study Group 2 Dataset , author =. 2025 , note =

    work page 2025

  72. [72]

    2025 , note =

    GBSG2: German Breast Cancer Study Group 2 , author =. 2025 , note =

    work page 2025

  73. [73]

    2025 , note =

    Rheumatoid Arthritis Data (ra\_data) , author =. 2025 , note =

    work page 2025

  74. [74]

    PLOS ONE , year =

    Disease activity trajectories for early and established rheumatoid arthritis: Real-world data from a rheumatoid arthritis cohort , author =. PLOS ONE , year =. doi:10.1371/journal.pone.0274264 , url =

    work page doi:10.1371/journal.pone.0274264

  75. [75]

    BMC Rheumatology , year =

    Using the derived 28-joint disease activity score patient-reported components (DAS28-P) index as a discriminatory measure of response to disease-modifying anti-rheumatic drug therapy in early rheumatoid arthritis , author =. BMC Rheumatology , year =. doi:10.1186/s41927-022-00299-3 , url =

    work page doi:10.1186/s41927-022-00299-3

  76. [76]

    2014 , howpublished =

    Bioequivalence Studies with Pharmacokinetic Endpoints for Drugs Submitted Under an ANDA: Guidance for Industry , author =. 2014 , howpublished =

    work page 2014

  77. [77]

    2010 , howpublished =

    Guideline on the Investigation of Bioequivalence , author =. 2010 , howpublished =

    work page 2010

  78. [78]

    2015 , howpublished =

    Scientific Considerations in Demonstrating Biosimilarity to a Reference Product: Guidance for Industry , author =. 2015 , howpublished =

    work page 2015

  79. [79]

    2014 , howpublished =

    Guideline on Similar Biological Medicinal Products , author =. 2014 , howpublished =

    work page 2014

  80. [80]

    , title =

    Schuirmann, Donald J. , title =. Journal of Pharmacokinetics and Biopharmaceutics , year =

    work page

Showing first 80 references.