pith. sign in

arxiv: 2408.16898 · v4 · submitted 2024-08-29 · 💰 econ.TH

Robust Robustness

Ian Ball , Deniz Kattwinkel This is my paper

Pith reviewed 2026-05-23 21:32 UTC · model grok-4.3

classification 💰 econ.TH
keywords robustnessmaxminambiguity setsmechanism designpayoff guaranteesweak topology
0
0 comments X

The pith

Maxmin-optimal mechanisms often produce payoff guarantees that are not robust to small changes in the ambiguity set.

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

The paper defines a payoff guarantee as robust when it continues to hold approximately for probability distributions close to the ambiguity set under the weak topology. It shows that many mechanisms chosen to maximize the worst-case payoff under standard maxmin criteria fail this test because they are built around degenerate worst-case priors. Some common ambiguity sets already possess a structural property that automatically makes every payoff guarantee derived from them robust. The paper proves that any ambiguity set can be expanded by a small amount to acquire this same structural property.

Core claim

A mechanism's payoff guarantee over an ambiguity set is robust if the guarantee is approximately satisfied at priors near the ambiguity set in the weak topology. Many maxmin-optimal mechanisms in the literature give payoff guarantees that are not robust. Such mechanisms are often tailored to degenerate worst-case priors, making them simple but fragile. Conversely, some commonly used ambiguity sets satisfy a structural property which ensures that every associated payoff guarantee is robust. Any ambiguity set can be slightly enriched to satisfy this property.

What carries the argument

The structural property of an ambiguity set that forces every associated maxmin payoff guarantee to remain approximately valid at nearby priors under the weak topology.

If this is right

  • Mechanisms built around degenerate worst-case priors become fragile once nearby priors are considered.
  • Ambiguity sets already satisfying the structural property automatically deliver robust guarantees for every mechanism they induce.
  • Any given ambiguity set admits a minimal enrichment that restores the structural property without large changes to the set itself.
  • Designers can therefore convert non-robust mechanisms into robust ones by adjusting the ambiguity set rather than redesigning the mechanism.

Where Pith is reading between the lines

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

  • The approach may change which mechanisms are selected once robustness is required, favoring those that perform well across a neighborhood rather than at a single point.
  • It raises the question of how to compute the minimal enrichment of a given ambiguity set in concrete applications.
  • The same distinction between fragile and robust guarantees could be applied to other decision criteria that rely on worst-case analysis.

Load-bearing premise

Defining robustness as approximate satisfaction of the guarantee at priors near the ambiguity set in the weak topology is the right way to capture stability of the guarantee.

What would settle it

An explicit maxmin-optimal mechanism whose payoff guarantee is violated by some prior outside the ambiguity set but arbitrarily close to it in the weak topology.

Figures

Figures reproduced from arXiv: 2408.16898 by Deniz Kattwinkel, Ian Ball.

Figure 1
Figure 1. Figure 1: Monopoly revenue maximization with median restriction (left column) and moment restrictions (right column). that Θ is a subset of Euclidean space. Let Π be a closed ambiguity set that contains only absolutely continuous priors. By Proposition 2.iv, the payoff guarantee from a social choice function f over Π is robust as long as the induced value function u ◦ f is continuous almost everywhere (with respect … view at source ↗
Figure 2
Figure 2. Figure 2: Monopoly regret-minimization with support restriction mechanism is a posted price, and Nature’s optimal prior has finite support, with a point mass at that posted price. The payoff guarantee from this mechanism is not robust. We show below (Theorem 1) that in problems with transfers, every saddle point with a finite-support prior yields a payoff guarantee that is either trivial or non-robust. By contrast, … view at source ↗
Figure 3
Figure 3. Figure 3: Bayesian persuasion: maxmin-optimal experiment (blue) and worst-case prior (orange) the designer’s utility u be negative ex-post regret.19 Let the ambiguity set Π contain all valuation distributions that concentrate on [ ¯ θ, 1]. Bergemann and Schlag (2008, Proposition 1, p. 565) solve for the unique maxmin-optimal social choice function ˆf. There are two cases. If ¯ θ ≤ 1/e, then ˆf is implemented by rand… view at source ↗
Figure 4
Figure 4. Figure 4: Delegated project choice: two-tiered mechanism (blue) and worst-case feasible set (orange) post for each good k a price pˆk ∈ argmaxp p · πk[p, ∞). The value function induced by this mechanism is continuous outside of the set ∪ K k=1{θ ∈ RK + : θk = ˆpk}. As long as πk({pˆk}) = 0 for each k, the payoff guarantee from this mechanism is robust (by Proposition 2.iv). Here, posting a single price for each good… view at source ↗
Figure 5
Figure 5. Figure 5: Robustified regret-minimization with support restriction where κ( ¯ θ, r) = ¯ θ( ¯ θe) −1/α( ¯ θ,r) ∈ (1/e, ¯ θ) and α( ¯ θ, r) ∈ [2,∞); the function α is defined in the proof. The associated regret guarantee is ¯ θ − α( ¯ θ, r)( ¯ θ − κ( ¯ θ, r)) + (α( ¯ θ, 1) − 1)r. In Bergemann and Schlag (2008), the maxmin-optimal mechanism’s regret guaran￾tee over Π is robust if and only if ¯ θ ≤ 1/e. If ¯ θ ≤ 1/e, th… view at source ↗
read the original abstract

We propose a refinement of the maxmin approach to robustness. A mechanism's payoff guarantee over an ambiguity set is \emph{robust} if the guarantee is approximately satisfied at priors near the ambiguity set (in the weak topology). We show that many maxmin-optimal mechanisms in the literature give payoff guarantees that are not robust. Such mechanisms are often tailored to degenerate worst-case priors, making them simple but fragile. Conversely, some commonly used ambiguity sets satisfy a structural property which ensures that every associated payoff guarantee is robust. We show how any ambiguity set can be slightly enriched to satisfy this property.

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 / 2 minor

Summary. The paper proposes a refinement of the maxmin approach to robustness in mechanism design under ambiguity. A mechanism's payoff guarantee over an ambiguity set is defined to be robust if the guarantee is approximately satisfied at priors near the ambiguity set in the weak topology. It shows that many maxmin-optimal mechanisms in the literature produce non-robust guarantees, often because they are tailored to degenerate worst-case priors. Conversely, some commonly used ambiguity sets satisfy a structural property ensuring that every associated payoff guarantee is robust, and the paper shows that any ambiguity set can be slightly enriched to satisfy this property.

Significance. If the results hold, the work offers a conceptually clean way to strengthen robustness guarantees in mechanism design so that they are not fragile to small perturbations around the ambiguity set. The structural property and the enrichment construction provide a general method that could be applied to existing ambiguity sets in auction, contract, and information design settings without requiring entirely new modeling frameworks. The emphasis on weak-topology continuity of the guarantee is a natural and falsifiable refinement that addresses a practical limitation of standard maxmin analysis.

major comments (2)
  1. [§3] §3 (Definition of robust payoff guarantee): the requirement that the guarantee holds approximately at all nearby priors in the weak topology is load-bearing for the subsequent claims about non-robustness of existing mechanisms; an explicit counter-example mechanism (with its ambiguity set and worst-case prior) showing failure of this continuity while satisfying the original maxmin criterion would strengthen the argument that the refinement is necessary.
  2. [§4] §4 (structural property and enrichment): the claim that the enrichment can be made 'slight' while preserving the original maxmin value needs to be stated with a precise metric on the space of ambiguity sets (e.g., Hausdorff distance or total-variation); without this, it is unclear whether the construction changes the set of optimal mechanisms or only the robustness property.
minor comments (2)
  1. [Introduction] The abstract and introduction would benefit from one concrete numerical or low-dimensional example (e.g., a simple auction or bilateral trade setting) illustrating both a non-robust mechanism and its enriched robust counterpart.
  2. [Notation] Notation for the ambiguity set Π and the set of nearby priors should be introduced once and used consistently; currently the weak-topology neighborhoods are described in prose without a displayed definition.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive assessment and the recommendation of minor revision. The comments help clarify the presentation of the robustness refinement and the enrichment construction. We address each point below and will incorporate the suggested clarifications.

read point-by-point responses

  1. Referee: [§3] §3 (Definition of robust payoff guarantee): the requirement that the guarantee holds approximately at all nearby priors in the weak topology is load-bearing for the subsequent claims about non-robustness of existing mechanisms; an explicit counter-example mechanism (with its ambiguity set and worst-case prior) showing failure of this continuity while satisfying the original maxmin criterion would strengthen the argument that the refinement is necessary.

    Authors: We agree that making the continuity failure fully explicit strengthens the motivation. The examples in Section 4 already demonstrate mechanisms whose maxmin value is achieved only at degenerate priors, so that the payoff guarantee drops discontinuously under weak convergence to nearby non-degenerate measures. In the revision we will add a self-contained, low-dimensional example (a simple posted-price mechanism with a two-point ambiguity set) that isolates the discontinuity while preserving the original maxmin optimality, thereby directly illustrating the load-bearing role of the continuity requirement. revision: yes

  2. Referee: [§4] §4 (structural property and enrichment): the claim that the enrichment can be made 'slight' while preserving the original maxmin value needs to be stated with a precise metric on the space of ambiguity sets (e.g., Hausdorff distance or total-variation); without this, it is unclear whether the construction changes the set of optimal mechanisms or only the robustness property.

    Authors: We will make the notion of 'slight' precise by equipping the space of compact ambiguity sets with the Hausdorff metric induced by the weak topology on probability measures. The enrichment construction adds, for each original prior, a small ball of measures whose radius can be chosen arbitrarily small; because the value function is continuous in the weak topology under the maintained assumptions, the maxmin value is unchanged for sufficiently small radius. Consequently the set of optimal mechanisms remains the same while the robustness property is restored. This metric and the continuity argument will be stated explicitly in the revised Section 5. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper introduces a new definition of robustness for payoff guarantees (approximate satisfaction at nearby priors in the weak topology), identifies a structural property on ambiguity sets that ensures robustness for all associated mechanisms, and constructs a slight enrichment of any ambiguity set to satisfy the property. These objects and the associated theorems are defined and proved directly from the stated primitives without any reduction of a claimed prediction or result to a fitted parameter, self-citation chain, or definitional equivalence. No load-bearing step relies on renaming a known result, smuggling an ansatz via citation, or invoking a uniqueness theorem from the authors' prior work. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper relies on standard concepts from decision theory and probability but introduces new definitions; no free parameters or invented entities are apparent from the abstract.

axioms (1)
  • domain assumption The weak topology is the appropriate metric for defining 'near' priors in the robustness refinement.
    Invoked directly in the definition of robust robustness in the abstract.

pith-pipeline@v0.9.0 · 5606 in / 1204 out tokens · 26852 ms · 2026-05-23T21:32:55.179714+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

51 extracted references · 51 canonical work pages

  1. [1]

    A Model of Delegated Project Choice,

    Aliprantis, C. D. and K. C. Border (2006): Infinite Dimensional Analysis: A Hitchhiker’s Guide, Springer, 3 ed. [8, 16, 37] Armstrong, M. and J. Vickers (2010): “A Model of Delegated Project Choice,” Econometrica, 78, 213–244

    work page 2006

  2. [2]

    Robust Contracting under Common Value Uncertainty,

    Auster, S. (2018): “Robust Contracting under Common Value Uncertainty,”Theo- retical Economics, 13, 175–204

    work page 2018

  3. [3]

    Optimal Parametric Auctions,

    Azar, P. and S. Micali (2012): “Optimal Parametric Auctions,” Computer Science and Artificial Intelligence Laboratory Technical Report MIT-CSAIL-TR-2012-011

    work page 2012

  4. [4]

    Distributional Robustness: From Pricing to Auctions,

    Bachrach, N. and I. Talgam-Cohen (2022): “Distributional Robustness: From Pricing to Auctions,”arXiv:2205.09008

    work page arXiv 2022

  5. [5]

    Correlation-Robust Analysis of Single Item Auction,

    Bei, X., N. Gra vin, P. Lu, and Z. G. Tang (2019): “Correlation-Robust Analysis of Single Item Auction,” inProceedings of the 2019 Annual ACM-SIAM Symposium on Discrete Algorithms (SODA), 193–208

    work page 2019

  6. [6]

    Robust Mechanism Design,

    Bergemann, D. and S. Morris (2005): “Robust Mechanism Design,”Economet- rica, 1771–1813

    work page 2005

  7. [7]

    An Introduction to Robust Mecha- nism Design,

    Bergemann, D., S. Morris, et al. (2013): “An Introduction to Robust Mecha- nism Design,”Foundations and Trends in Microeconomics, 8, 169–230

    work page 2013

  8. [8]

    Robust Monopoly Pricing,

    Bergemann, D. and K. Schlag (2011): “Robust Monopoly Pricing,”Journal of Economic Theory, 146, 2527–2543

    work page 2011

  9. [9]

    Pricing without Priors,

    Bergemann, D. and K. H. Schlag (2008): “Pricing without Priors,”Journal of the European Economic Association, 6, 560–569. [3, 12, 13, 22, 28, 29, 45] Billingsley, P. (1999): Convergence of Probability Measures, John Wiley & Sons, 2 ed

    work page 2008

  10. [10]

    Maxmin Auction Design with Known Expected Values,

    Bogachev, V. I. (2018): Weak Convergence of Measures, vol. 234 ofMathematical Surveys and Monographs, American Mathematical Society. [25, 43] Brooks, B. and S. Du (2021a): “Maxmin Auction Design with Known Expected Values,” Technical report, University of Chicago and University of California–San Diego

    work page 2018

  11. [11]

    Robust Mechanisms for the Financing of Public Goods,

    ——— (2023): “Robust Mechanisms for the Financing of Public Goods,”Available at 49 SSRN 4482541

    work page 2023

  12. [12]

    Optimal Selling Mechanisms under Moment Conditions,

    Carrasco, V., V. F arinha Luz, N. Kos, M. Messner, P. Monteiro, and H. Moreira (2018): “Optimal Selling Mechanisms under Moment Conditions,” Journal of Economic Theory, 177, 245–279. [11, 24] Carroll, G. (2015): “Robustness and Linear Contracts,”American Economic Re- view, 105, 536–63. [17, 18, 53, 54, 55] ——— (2017): “Robustness and Separation in Multidi...

    work page 2018

  13. [13]

    Continuous Implementation with Payoff Knowledge,

    Chen, Y.-C., T. Kunimoto, and Y. Sun (2023): “Continuous Implementation with Payoff Knowledge,”Journal of Economic Theory, 209, 105624

    work page 2023

  14. [14]

    Continuous Imple- mentation with Direct Revelation Mechanisms,

    Chen, Y.-C., M. Mueller-Frank, and M. M. Pai (2022): “Continuous Imple- mentation with Direct Revelation Mechanisms,”Journal of Economic Theory, 201, 105422

    work page 2022

  15. [15]

    Foundations of Dominant-Strategy Mecha- nisms,

    Chung, K.-S. and J. C. Ely (2007): “Foundations of Dominant-Strategy Mecha- nisms,”Review of Economic Studies, 74, 447–476

    work page 2007

  16. [16]

    Strong Duality for a Multiple-Good Monopolist,

    Daskalakis, C., A. Deckelbaum, and C. Tzamos (2017): “Strong Duality for a Multiple-Good Monopolist,”Econometrica, 85, 735–767

    work page 2017

  17. [17]

    Robust Mechanisms Under Common Valuation,

    Du, S. (2018): “Robust Mechanisms Under Common Valuation,”Econometrica, 86, 1569–1588

    work page 2018

  18. [18]

    Robust Contracting for Sequen- tial Search,

    Durandard, T., U. V aidya, and B. Xu (2025): “Robust Contracting for Sequen- tial Search,”arXiv:2504.17948

    work page arXiv 2025

  19. [19]

    Preparing for the Worst but Hoping for the Best: Robust (Bayesian) Persuasion,

    Dworczak, P. and A. Pa v an(2022): “Preparing for the Worst but Hoping for the Best: Robust (Bayesian) Persuasion,”Econometrica, 90, 2017–2051

    work page 2022

  20. [20]

    (2008): Measure, Topology, and Fractal Geometry, Springer, 2 ed

    Edgar, G. (2008): Measure, Topology, and Fractal Geometry, Springer, 2 ed

    work page 2008

  21. [21]

    Maxmin Expected Utility with Non- Unique Prior,

    Gilboa, I. and D. Schmeidler (1989): “Maxmin Expected Utility with Non- Unique Prior,”Journal of Mathematical Economics, 18, 141–153. [5, 6, 7, 34] Guo, Y. and E. Shmaya (2023): “Regret-Minimizing Project Choice,”Economet- rica, 91, 1567–1593. [15, 16] Hansen, L. P. and T. J. Sargent (2001): “Robust Control and Model Uncer- 50 tainty,”American Economic Re...

    work page 1989

  22. [22]

    Structured Ambiguity and Model Misspecification,

    ——— (2022): “Structured Ambiguity and Model Misspecification,”Journal of Eco- nomic Theory, 199, 105165

    work page 2022

  23. [23]

    RobustControlandModelMisspecification,

    Hansen, L. P., T. J. Sargent, G. Turmuhambetov a, and N. Williams (2006): “RobustControlandModelMisspecification,”Journal of Economic Theory, 128, 45–90

    work page 2006

  24. [24]

    Correlation-Robust Auction Design,

    He, W. and J. Li (2022): “Correlation-Robust Auction Design,”Journal of Eco- nomic Theory, 200, 105403

    work page 2022

  25. [25]

    ANon-BayesianTheoryofState-DependentUtility,

    Hill, B. (2019): “ANon-BayesianTheoryofState-DependentUtility,”Econometrica, 87, 1341–1366

    work page 2019

  26. [26]

    Robust Pricing with Refunds,

    Hinnosaar, T. and K. Ka w ai (2020): “Robust Pricing with Refunds,”RAND Journal of Economics, 51, 1014–1036

    work page 2020

  27. [27]

    Robust Persuasion of a Privately Informed Receiver,

    Hu, J. and X. Weng (2021): “Robust Persuasion of a Privately Informed Receiver,” Economic Theory, 72, 909–953. [13, 14] Jehiel, P., M. Meyer-ter Vehn, and B. Moldov anu (2012): “Locally Robust Implementation and its Limits,”Journal of Economic Theory, 147, 2439–2452

    work page 2021

  28. [28]

    Randomization and the Robustness of Linear Contracts,

    Kambhampati, A., B. Peng, Z. G. Tang, J. Toikka, and R. Vohra (2025): “Randomization and the Robustness of Linear Contracts,” Working paper

    work page 2025

  29. [29]

    Bayesian Persuasion,

    Kamenica, E. and M. Gentzkow (2011): “Bayesian Persuasion,”American Eco- nomic Review, 101, 2590–2615. [13, 14] Kleiner, A. and A. Manelli (2019): “Strong Duality in Monopoly Pricing,” Econometrica, 87, 1391–1396

    work page 2011

  30. [30]

    Persuasion with Unknown Beliefs,

    Kosterina, S. (2022): “Persuasion with Unknown Beliefs,”Theoretical Economics, 17, 1075–1107

    work page 2022

  31. [31]

    Continuity of Monotone Functions,

    La vrič, B.(1993): “Continuity of Monotone Functions,”Archivum Mathematicum, 29, 1–4

    work page 1993

  32. [32]

    Informational Robustness in Intertemporal Pric- ing,

    Libgober, J. and X. Mu (2021): “Informational Robustness in Intertemporal Pric- ing,”Review of Economic Studies, 88, 1224–1252

    work page 2021

  33. [33]

    Ambiguity Aver- sion, Robustness, and the Variational Representation of Preferences,

    Maccheroni, F., M. Marinacci, and A. Rustichini (2006): “Ambiguity Aver- sion, Robustness, and the Variational Representation of Preferences,”Economet- rica, 74, 1447–1498. [3, 7, 27, 34] Madarász, K. and A. Prat (2017): “Sellers with Misspecified Models,”Review of 51 Economic Studies, 84, 790–815

    work page 2006

  34. [34]

    The Robustness of Robust Imple- mentation,

    Meyer-ter Vehn, M. and S. Morris (2011): “The Robustness of Robust Imple- mentation,”Journal of Economic Theory, 146, 2093–2104

    work page 2011

  35. [35]

    ContinuousImplementationwithLocalPayoffUncertainty,

    Oury, M.(2015): “ContinuousImplementationwithLocalPayoffUncertainty,”Jour- nal of Economic Theory, 159, 656–677

    work page 2015

  36. [36]

    Continuous Implementation,

    Oury, M. and O. Tercieux (2012): “Continuous Implementation,”Econometrica, 80, 1605–1637

    work page 2012

  37. [37]

    Robust Implementation with Costly Informa- tion,

    Pei, H. and B. Strulovici (2024): “Robust Implementation with Costly Informa- tion,”Review of Economic Studies

    work page 2024

  38. [38]

    The Cost of Information: The Case of Constant Marginal Costs,

    Pomatto, L., P. Strack, and O. Tamuz (2023): “The Cost of Information: The Case of Constant Marginal Costs,”American Economic Review, 113, 1360–1393

    work page 2023

  39. [39]

    Non-Robustness of Some Economic Models,

    Prasad, K. (2003): “Non-Robustness of Some Economic Models,”Topics in Theo- retical Economics,

    work page 2003

  40. [40]

    Frameworks and Results in Distribu- tionally Robust Optimization,

    Rahimian, H. and S. Mehrotra (2022): “Frameworks and Results in Distribu- tionally Robust Optimization,”Open Journal of Mathematical Optimization,

    work page 2022

  41. [41]

    A Min-max Solution of an Inventory Problem,

    Scarf, H. (1958): “A Min-max Solution of an Inventory Problem,” inStudies in the Mathematical Theory of Inventory and Production, ed. by K. Arrow, S. Karlin, and H. Scarf, Stanford University Press, chap. 12, 201–209

    work page 1958

  42. [42]

    Robust Bayesian Choice,

    Stanca, L. (2023): “Robust Bayesian Choice,”Mathematical Social Sciences, 126, 94–106

    work page 2023

  43. [43]

    (2009): Optimal Transport: Old and New, Springer-Verlag

    Villani, C. (2009): Optimal Transport: Old and New, Springer-Verlag

    work page 2009

  44. [44]

    Optimal Search for the Best Alternative,

    Weitzman, M. L. (1979): “Optimal Search for the Best Alternative,”Econometrica, 47, 641–654

    work page 1979

  45. [45]

    Correlation-Robust Optimal Auctions,

    Zhang, W. (2022): “Correlation-Robust Optimal Auctions,”arXiv:2105.04697

    work page arXiv 2022

  46. [46]

    To make this robustness property stronger, we consider a largerstatespacethatallowsforadditionaloutputperturbations

    52 B Online appendix: Additional results B.1 Robustness of payoff guarantee in Carroll (2015) In the setting of Carroll (2015), we prove that the payoff guarantee from each nonzero linear contract is robust. To make this robustness property stronger, we consider a largerstatespacethatallowsforadditionaloutputperturbations. Recallthat Y ⊂ R+ is a compact s...

    work page 2015

  47. [47]

    In particular, Y ′ can contain negative output levels

    Let Y ′ be a compact subset of R satisfying Y ′ ⊇ Y . In particular, Y ′ can contain negative output levels. Here, a technology is a nonempty compact subset of∆(Y ′) × R. Define the state spaceΘ to be the collection of all technologies. We now define a metric onΘ. First, endow Y ′ with the usual absolute value metric, and letdW denote the associated Wasse...

    work page 2015

  48. [48]

    The calculations in Carroll (2015, p

    work page 2015

  49. [49]

    57 • Given priorsµ, ν ∈ ∆(Θ), writeµ ≪ ν if µ is absolutely continuous with respect to ν

    • A support–moment setis defined by M(S; g, Y ) = {π ∈ ∆(Θ) : π(S) = 1 and Eθ∼π[g(θ)] ∈ Y } , for some measurable proper subsetS of Θ, some continuous functiong : Θ → Rm, and some subsetY of Rm that intersects the relative interior ofconv g(S). 57 • Given priorsµ, ν ∈ ∆(Θ), writeµ ≪ ν if µ is absolutely continuous with respect to ν. The relative entropy(K...

    work page 2001

  50. [50]

    60), the sequences (πn) and (π′ n) are both tight

    By Prokhorov’s theorem (Billingsley, 1999, Theorem 5.2, p. 60), the sequences (πn) and (π′ n) are both tight. Thus, there exists a compact set K such that for all n we have πn(K) ≥ 1 − ε and π′ n(K) ≥ 1 − ε.69 The function H, being continuous, must achieve a maximum over the compact setK. Let L = 1 + maxθ∈K H(θ). Let C = {θ ∈ Θ : H(θ) ≥ L}. The setC is cl...

    work page 1999

  51. [51]

    Select a bounded, compatible metric d

    Let (θj) be an enumeration of N. Select a bounded, compatible metric d. For each n, let πn = ∞X j=1 π B(θj, 1/n) \ ∪j−1 k=1B(θk, 1/n) δθj . Using Tonelli’s theorem, it can be shown thatπn is countably additive and hence a 65 probability measure. By construction, W (πn, π) ≤ 1/n, so the sequence (πn) con- verges to π in the Wasserstein metric, and hence we...

    work page 2009