pith. sign in

arxiv: 2112.13247 · v3 · submitted 2021-12-25 · 🧮 math.ST · stat.TH

Decision-making with possibilistic inferential models

Ryan Martin , Shih-Ni Prim , Jonathan Williams This is my paper

Pith reviewed 2026-05-24 12:58 UTC · model grok-4.3

classification 🧮 math.ST stat.TH
keywords possibilistic inferential modelsdecision makingChoquet integralimprecise probabilityreliabilityoracle comparisonequivariant models
0
0 comments X

The pith

Possibilistic inferential models yield reliable decisions because their Choquet integral action assessments are not overly optimistic relative to an oracle.

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

The paper extends likelihood-based possibilistic inferential models from uncertainty quantification in inference to a corresponding framework for decision making. It establishes that the model's assessment of an action's quality, computed with a simple Choquet integral, tends not to be too optimistic compared to an oracle's assessment. This property ensures the model does not favor actions that an oracle would reject. A large-sample efficiency result shows the reliability is not achieved through excessive conservatism. In the special case of equivariant statistical models, connections between the IM's and Bayesian recommended actions support certain optimality conclusions.

Core claim

Focusing on a likelihood-based possibilistic IM formulation, the framework for decision making assesses an action's quality by a simple Choquet integral. This assessment tends not to be too optimistic compared to that of an oracle. This ensures that the IM tends not to favor actions that the oracle does not also favor, hence the IM is also reliable for decision making. A complementary large-sample efficiency result establishes that this reliability is not achieved by being grossly conservative. In the special case of equivariant statistical models, further connections can be made between the IM's and Bayesian's recommended actions, from which certain optimality conclusions can be drawn.

What carries the argument

Choquet integral used to assess action quality within the likelihood-based possibilistic inferential model

If this is right

  • The IM tends not to favor actions that the oracle does not also favor.
  • Reliability for decision making is complemented by large-sample efficiency that rules out gross conservatism.
  • In equivariant statistical models, links to Bayesian recommended actions support optimality conclusions.

Where Pith is reading between the lines

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

  • The same Choquet-based assessment might be tested in sequential decision problems where data arrives over time.
  • Connections to Bayesian actions in the equivariant case suggest possible hybrid rules that blend the two approaches in symmetric settings.

Load-bearing premise

The reliability guarantees of the likelihood-based possibilistic IM for inference extend to the decision-making setting through the use of the Choquet integral for action assessment.

What would settle it

A concrete statistical model and data set in which the IM's Choquet integral assessment assigns higher quality to an action that yields strictly lower expected payoff under the true parameter than an alternative action preferred by the oracle.

Figures

Figures reproduced from arXiv: 2112.13247 by Jonathan Williams, Ryan Martin, Shih-Ni Prim.

Figure 1
Figure 1. Figure 1: Plot of the possibility contour πy(ϑ), for y = 0, when PU is a Student-t distribution with 3 degrees of freedom; dashed line shows the loss `a(ϑ), when a = 2. Next, by the extension principle of Zadeh (1978), for the observed data Y = y, the corresponding possibility contour on the parameter space is given by πy(ϑ) = π(y − ϑ), ϑ ∈ R. As this is the first example, to keep things sufficiently simple that the… view at source ↗
Figure 2
Figure 2. Figure 2: The black line shows the distribution function of [PITH_FULL_IMAGE:figures/full_fig_p018_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Skew-normal example with y = 0. Panel (a) shows the possibility contour πy(ϑ) for the skew-normal illustration; overlaid is the squared error loss function `a(·) with a = −2. Panel (b) shows the IM (solid) and fiducial (dashed) expected losses. 4.3.2 Binomial model Consider a binomial model, PY |θ = Bin(n, θ), where the number of trials, n, is known but the success probability, θ ∈ [0, 1], is unknown. Ther… view at source ↗
Figure 4
Figure 4. Figure 4: Binomial example with n = 18 and y = 7. Panel (a) shows the possibility contour πy(ϑ); dashed line shows the weighted squared error loss function `a(ϑ), when a = 0.2. Panel (b) shows the (upper) expected loss functions, a 7→ Πy`a (solid) and a 7→ Q? y `a (dashed), for the weighted squared error loss function. This loss function is overlaid in [PITH_FULL_IMAGE:figures/full_fig_p020_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The black line shows the distribution function of [PITH_FULL_IMAGE:figures/full_fig_p021_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Same possibility contour πy(ϑ) (solid) as in [PITH_FULL_IMAGE:figures/full_fig_p027_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Contour plot showing level sets of πy (gray) and of `a (dotted). The solid black line is the path α 7→ θα, which is a curve in this example. The challenge to extending the above convexity argument to more than one dimension is that the path α 7→ θα is generally not a line [PITH_FULL_IMAGE:figures/full_fig_p030_7.png] view at source ↗
read the original abstract

Inferential models (IMs) are data-dependent, imprecise-probabilistic structures designed to quantify uncertainty about unknowns. As the name suggests, the focus has been on uncertainty quantification for inference and on its reliability properties in that context. Focusing on a likelihood-based possibilistic IM formulation, the present paper develops a corresponding framework for decision making, and investigates the decision-theoretic implications of the IM's reliability guarantees. Here we show that the possibilistic IM's assessment of an action's quality, defined by a simple Choquet integral, tends not be too optimistic compared to that of an oracle. This ensures that the IM tends not to favor actions that the oracle doesn't also favor, hence the IM is also reliable for decision making. We also establish a complementary, large-sample efficiency result that says the IM's reliability isn't achieved by being grossly conservative. In the special case of equivariant statistical models, further connections can be made between the IM's and Bayesian's recommended actions, from which certain optimality conclusions can be drawn.

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

0 major / 3 minor

Summary. The manuscript develops a decision-making framework based on likelihood-based possibilistic inferential models (IMs). It shows that an action's quality, assessed via a simple Choquet integral, is not too optimistic relative to an oracle's assessment; this prevents the IM from favoring actions the oracle would reject, thereby extending the IM's reliability guarantees from inference to decisions. A complementary large-sample efficiency result is established to rule out trivial conservatism, and in the special case of equivariant models further connections to Bayesian recommendations and optimality conclusions are drawn.

Significance. If the derivations hold, the work supplies a direct theoretical bridge from the reliability properties of possibilistic IMs to decision theory, furnishing non-optimism and efficiency guarantees that are not achieved by ad-hoc conservatism. The equivariant-model results add concrete optimality links that may be of independent interest to both imprecise-probability and classical decision theorists.

minor comments (3)
  1. [§2.2] §2.2: the notation for the lower and upper probabilities induced by the possibilistic IM is introduced without an explicit cross-reference to the earlier definition in Eq. (3); adding the pointer would improve readability.
  2. [§4] §4, Theorem 2: the statement of the large-sample efficiency result would be clearer if the precise rate (e.g., o_p(1) or O_p(n^{-1/2})) were written explicitly rather than left implicit in the surrounding prose.
  3. [Figure 1] Figure 1: the caption does not indicate whether the plotted contours are for a fixed sample size or averaged over replications; a brief clarification would help readers interpret the numerical illustration.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary of our manuscript and the recommendation of minor revision. The assessment that the work provides a theoretical bridge from IM reliability to decision theory is appreciated.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The provided abstract and context describe an extension of existing IM reliability properties to a decision-making setting via a Choquet integral construction, plus a separate large-sample efficiency result. No equations, self-citations, fitted parameters, or ansatzes are quoted that reduce any central claim to its own inputs by construction. The derivation chain is presented as carrying over the inference guarantees directly through the integral definition without evidence of self-definition or renaming. This is the expected honest non-finding when the text supplies no load-bearing reduction to quote.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract does not provide enough information to identify any free parameters, axioms, or invented entities; the work builds on existing inferential model concepts.

pith-pipeline@v0.9.0 · 5699 in / 1057 out tokens · 35289 ms · 2026-05-24T12:58:37.766450+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Regularized e-processes: anytime valid inference with knowledge-based efficiency gains

    math.ST 2024-10 unverdicted novelty 6.0

    Regularized e-processes add knowledge-based imprecise-probabilistic regularization to e-processes, yielding anytime-valid inference with efficiency gains and possibility-theoretic uncertainty quantification that satis...

  2. Possibilistic inferential models: a review

    math.ST 2025-07 unverdicted novelty 4.0

    A review of possibilistic inferential models that deliver strong frequentist reliability and conditional imprecise-probabilistic reasoning, plus a generalization connecting them to bootstrap and conformal prediction methods.

Reference graph

Works this paper leans on

70 extracted references · 70 canonical work pages · cited by 2 Pith papers · 1 internal anchor

  1. [1]

    Augustin, T., Walter, G., and Coolen, F. P. A. (2014). Statistical inference. In Introduction to I mprecise P robabilities , Wiley Ser. Probab. Stat., pages 135--189. Wiley, Chichester

    work page 2014

  2. [2]

    Azzalini, A. (1985). A class of distributions which includes the normal ones. Scand. J. Statist. , 12(2):171--178

    work page 1985

  3. [3]

    Balch, M. S. (2012). Mathematical foundations for a theory of confidence structures. Internat. J. Approx. Reason. , 53(7):1003--1019

    work page 2012

  4. [4]

    Balch, M. S. (2020). New two-sided confidence intervals for binomial inference derived using W alley's imprecise posterior likelihood as a test statistic. Internat. J. Approx. Reason. , 123:77--98

    work page 2020

  5. [5]

    S., Martin, R., and Ferson, S

    Balch, M. S., Martin, R., and Ferson, S. (2019). Satellite conjunction analysis and the false confidence theorem. Proc. Royal Soc. A , 475(2227):2018.0565

    work page arXiv 2019

  6. [6]

    Berger, J. O. (1984). The robust B ayesian viewpoint. In Robustness of B ayesian A nalyses , volume 4 of Stud. Bayesian Econometrics , pages 63--144. North-Holland, Amsterdam. With comments and with a reply by the author

    work page 1984

  7. [7]

    Berger, J. O. (1985). Statistical Decision Theory and B ayesian Analysis . Springer-Verlag, New York, second edition

    work page 1985

  8. [8]

    Bergmann, R. (1991). Stochastic orders and their application to a unified approach to various concepts of dependence and association. In Stochastic Orders and Decision Under Risk ( H amburg, 1989) , volume 19 of IMS Lecture Notes Monogr. Ser. , pages 48--73. Inst. Math. Statist., Hayward, CA

    work page 1991

  9. [9]

    Bernoulli, D. (1954). Exposition of a new theory on the measurement of risk. Econometrica , 22:23--36

    work page 1954

  10. [10]

    Brown, L. D. (1971). Admissible estimators, recurrent diffusions, and insoluble boundary value problems. Ann. Math. Statist. , 42:855--903

    work page 1971

  11. [11]

    and Martin, R

    Cahoon, J. and Martin, R. (2021). Generalized inferential models for censored data. Internat. J. Approx. Reason. , 137:51--66

    work page 2021

  12. [12]

    and Martin, R

    Cella, L. and Martin, R. (2021a). Valid inferential models for prediction in supervised learning problems. In Cano, A., De Bock, J., Miranda, E., and Moral, S., editors, Proceedings of the Twelveth International Symposium on Imprecise Probability: Theories and Applications , volume 147 of Proceedings of Machine Learning Research , pages 72--82. PMLR. Exte...

    work page

  13. [13]

    and Martin, R

    Cella, L. and Martin, R. (2021b). Validity, consonant plausibility measures, and conformal prediction. Internat. J. Approx. Reason. , to appear; https://researchers.one/articles/20.01.00010

    work page

  14. [14]

    (1953--1954)

    Choquet, G. (1953--1954). Theory of capacities. Ann. Inst. Fourier, Grenoble , 5:131--295 (1955)

    work page 1953

  15. [15]

    Crane, H. (2018). The fundamental principle of probability: R esolving the replication crisis with skin in the game. Researchers.One , https://researchers.one/articles/18.08.00013

    work page 2018

  16. [16]

    Cuzzolin, F. (2021). The Geometry of Uncertainty . Artificial Intelligence: Foundations, Theory, and Algorithms. Springer, Cham

    work page 2021

  17. [17]

    Dawid, A. P. and Stone, M. (1982). The functional-model basis of fiducial inference. Ann. Statist. , 10(4):1054--1074. With discussion

    work page 1982

  18. [18]

    Dempster, A. P. (1963). Further examples of inconsistencies in the fiducial argument. Ann. Math. Statist. , 34:884--891

    work page 1963

  19. [19]

    Dempster, A. P. (2008). The D empster-- S hafer calculus for statisticians. Internat. J. Approx. Reason. , 48(2):365--377

    work page 2008

  20. [20]

    Den ux, T. (2019). Decision-making with belief functions: a review. Internat. J. Approx. Reason. , 109:87--110

    work page 2019

  21. [21]

    and Li, S

    Den ux, T. and Li, S. (2018). Frequency-calibrated belief functions: review and new insights. Internat. J. Approx. Reason. , 92:232--254

    work page 2018

  22. [22]

    and Dubois, D

    Destercke, S. and Dubois, D. (2014). Special cases. In Introduction to I mprecise P robabilities , Wiley Ser. Probab. Stat., pages 79--92. Wiley, Chichester

    work page 2014

  23. [23]

    Dubois, D. (2006). Possibility theory and statistical reasoning. Comput. Statist. Data Anal. , 51(1):47--69

    work page 2006

  24. [24]

    Dubois, D., Foulloy, L., Mauris, G., and Prade, H. (2004). Probability-possibility transformations, triangular fuzzy sets, and probabilistic inequalities. Reliab. Comput. , 10(4):273--297

    work page 2004

  25. [25]

    and Prade, H

    Dubois, D. and Prade, H. (1988). Possibility T heory . Plenum Press, New York

    work page 1988

  26. [26]

    Fisher, R. A. (1973). Statistical Methods and Scientific Inference . Hafner Press, New York, 3rd edition

    work page 1973

  27. [27]

    Fraser, D. A. S. (1968). The S tructure of I nference . John Wiley & Sons Inc., New York

    work page 1968

  28. [28]

    Fraser, D. A. S. (2011). Is B ayes posterior just quick and dirty confidence? Statist. Sci. , 26(3):299--316

    work page 2011

  29. [29]

    Fraser, D. A. S. (2013). Discussion: `` C onfidence distribution, the frequentist distribution estimator of a parameter: a review'' [mr3047496]. Int. Stat. Rev. , 81(1):42--48

    work page 2013

  30. [30]

    K., Delampady, M., and Samanta, T

    Ghosh, J. K., Delampady, M., and Samanta, T. (2006). An I ntroduction to B ayesian A nalysis . Springer, New York

    work page 2006

  31. [31]

    Giang, P. H. and Shenoy, P. P. (2011). A decision theory for partially consonant belief functions. Internat. J. Approx. Reason. , 52(3):375--394

    work page 2011

  32. [32]

    Gilboa, I. (1987). Expected utility with purely subjective nonadditive probabilities. J. Math. Econom. , 16(1):65--88

    work page 1987

  33. [33]

    and Schmeidler, D

    Gilboa, I. and Schmeidler, D. (1989). Maxmin expected utility with nonunique prior. J. Math. Econom. , 18(2):141--153

    work page 1989

  34. [34]

    and Schmeidler, D

    Gilboa, I. and Schmeidler, D. (1994). Additive representations of non-additive measures and the C hoquet integral. Ann. Oper. Res. , 52:43--65

    work page 1994

  35. [35]

    Gr\"unwald, P. (2018). Safe probability. J. Statist. Plann. Inference , 195:47--63

    work page 2018

  36. [36]

    Hannig, J., Iyer, H., Lai, R. C. S., and Lee, T. C. M. (2016). Generalized fiducial inference: a review and new results. J. Amer. Statist. Assoc. , 111(515):1346--1361

    work page 2016

  37. [37]

    and Hanss, M

    Hose, D. and Hanss, M. (2021). A universal approach to imprecise probabilities in possibility theory. Internat. J. Approx. Reason. , 133:133--158

    work page 2021

  38. [38]

    Huntley, N., Hable, R., and Troffaes, M. C. M. (2014). Decision making. In Introduction to I mprecise P robabilities , Wiley Ser. Probab. Stat., pages 190--206. Wiley, Chichester

    work page 2014

  39. [39]

    and Martin, R

    Liu, C. and Martin, R. (2021). Inferential models and possibility measures. Handbook of Bayesian, Fiducial, and Frequentist Inference , to appear; arXiv:2008.06874

    work page arXiv 2021

  40. [40]

    Martin, R. (2019). False confidence, non-additive beliefs, and valid statistical inference. Internat. J. Approx. Reason. , 113:39--73

    work page 2019

  41. [41]

    Martin, R. (2021a). An imprecise-probabilistic characterization of frequentist statistical inference. https://researchers.one/articles/21.01.00002

    work page

  42. [42]

    Martin, R. (2021b). Towards a theory of valid inferential models under partial prior information. https://researchers.one/articles/21.05.00001

    work page

  43. [43]

    Martin, R., Balch, M., and Ferson, S. (2021). Response to the comment ` C onfidence in confidence distributions!'. Proc. R. Soc. A. , 477:20200579

    work page 2021

  44. [44]

    Martin, R., Ermini Leaf, D., and Liu, C. (2012). Optimal inferential models for a P oisson mean. arXiv:1207.0105

    work page internal anchor Pith review Pith/arXiv arXiv 2012

  45. [45]

    and Liu, C

    Martin, R. and Liu, C. (2013). Inferential models: a framework for prior-free posterior probabilistic inference. J. Amer. Statist. Assoc. , 108(501):301--313

    work page 2013

  46. [46]

    and Liu, C

    Martin, R. and Liu, C. (2014). A note on p-values interpreted as plausibilities. Statist. Sinica , 24(4):1703--1716

    work page 2014

  47. [47]

    and Liu, C

    Martin, R. and Liu, C. (2015a). Conditional inferential models: combining information for prior-free probabilistic inference. J. R. Stat. Soc. Ser. B. Stat. Methodol. , 77(1):195--217

    work page

  48. [48]

    and Liu, C

    Martin, R. and Liu, C. (2015b). Inferential M odels: R easoning with U ncertainty , volume 147 of Monographs on Statistics and Applied Probability . CRC Press, Boca Raton, FL

    work page

  49. [49]

    Molchanov, I. (2005). Theory of Random Sets . Probability and Its Applications (New York). Springer-Verlag London Ltd., London

    work page 2005

  50. [50]

    Murph, A., Hannig, J., and Williams, J. P. (2021). Introduction to generalized fiducial inference. Handbook of Bayesian, Fiducial, and Frequentist Inference , to appear

    work page 2021

  51. [51]

    Nguyen, H. T. (2006). An Introduction to Random Sets . Chapman & Hall/CRC, Boca Raton, FL

    work page 2006

  52. [52]

    Savage, L. J. (1972). The Foundations of Statistics . Dover Publications, Inc., New York, revised edition

    work page 1972

  53. [53]

    and Hjort, N

    Schweder, T. and Hjort, N. L. (2002). Confidence and likelihood. Scand. J. Statist. , 29(2):309--332

    work page 2002

  54. [54]

    and Hjort, N

    Schweder, T. and Hjort, N. L. (2016). Confidence, L ikelihood, P robability , volume 41 of Cambridge Series in Statistical and Probabilistic Mathematics . Cambridge University Press, New York

    work page 2016

  55. [55]

    Shafer, G. (1976). A Mathematical Theory of Evidence . Princeton University Press, Princeton, N.J

    work page 1976

  56. [56]

    Shafer, G. (2021). Testing by betting: a strategy for statistical and scientific communication. J. Roy. Statist. Soc. Ser. A , 184(2):407--478

    work page 2021

  57. [57]

    Stein, C. (1956). Inadmissibility of the usual estimator for the mean of a multivariate normal distribution. In Proceedings of the T hird B erkeley S ymposium on M athematical S tatistics and P robability, 1954--1955, vol. I , pages 197--206, Berkeley and Los Angeles. University of California Press

    work page 1956

  58. [58]

    Taraldsen, G. (2021). Joint confidence distributions. https://doi.org/10.13140/RG.2.2.33079.85920

    work page doi:10.13140/rg.2.2.33079.85920 2021

  59. [59]

    and Lindqvist, B

    Taraldsen, G. and Lindqvist, B. H. (2013). Fiducial theory and optimal inference. Ann. Statist. , 41(1):323--341

    work page 2013

  60. [60]

    Troffaes, M. C. M. and de Cooman, G. (2014). Lower P revisions . Wiley Series in Probability and Statistics. John Wiley & Sons, Ltd., Chichester

    work page 2014

  61. [61]

    Troffaes, M. C. M. and Hable, R. (2014). Computation. In Introduction to I mprecise P robabilities , Wiley Ser. Probab. Stat., pages 329--337. Wiley, Chichester

    work page 2014

  62. [62]

    van der Vaart, A. (2002). The statistical work of L ucien L e C am. Ann. Statist. , 30(3):631--682. Dedicated to the memory of Lucien Le Cam

    work page 2002

  63. [63]

    Vidakovic, B. (2000). -minimax: a paradigm for conservative robust B ayesians. In Robust B ayesian analysis , volume 152 of Lect. Notes Stat. , pages 241--259. Springer, New York

    work page 2000

  64. [64]

    and Morgenstern, O

    von Neumann, J. and Morgenstern, O. (1947). Theory of G ames and E conomic B ehavior . Princeton University Press, Princeton, N. J. 2d ed

    work page 1947

  65. [65]

    Walley, P. (1987). Belief function representations of statistical evidence. Ann. Statist. , 15(4):1439--1465

    work page 1987

  66. [66]

    Walley, P. (1991). Statistical R easoning with I mprecise P robabilities , volume 42 of Monographs on Statistics and Applied Probability . Chapman & Hall Ltd., London

    work page 1991

  67. [67]

    Wang, R.-S. (2011). Some inequalities and convergence theorems for C hoquet integrals. J. Appl. Math. Comput. , 35(1-2):305--321

    work page 2011

  68. [68]

    Wasserman, L. A. (1990). Belief functions and statistical inference. Canad. J. Statist. , 18(3):183--196

    work page 1990

  69. [69]

    and Singh, K

    Xie, M. and Singh, K. (2013). Confidence distribution, the frequentist distribution estimator of a parameter: a review. Int. Stat. Rev. , 81(1):3--39

    work page 2013

  70. [70]

    Zadeh, L. A. (1978). Fuzzy sets as a basis for a theory of possibility. Fuzzy Sets and Systems , 1(1):3--28

    work page 1978