pith. machine review for the scientific record. sign in

arxiv: 2604.22667 · v1 · submitted 2026-04-24 · 🧮 math.ST · math.PR· stat.TH

Recognition: unknown

Sharp bounds for products of dependent random variables

Christopher Blier-Wong , Jinghui Chen
Authors on Pith no claims yet

Pith reviewed 2026-05-08 09:13 UTC · model grok-4.3

classification 🧮 math.ST math.PRstat.TH
keywords sharp boundsproduct expectationdependent random variablescomonotonic couplingparity polytopessign-bias vectorextremal copulas
0
0 comments X

The pith

Sharp bounds on the expected product of random variables with fixed marginals are attained exactly when the sign-bias vector lies in the even-parity polytope for the upper bound or the odd-parity polytope for the lower bound.

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

The paper determines the sharpest possible upper and lower bounds on the expectation of the product of d random variables when only the univariate marginal distributions are known. It decomposes the product into its magnitude part, handled by comonotonic coupling of the absolute values, and its sign part, governed by a sign-bias vector. Universal bounds follow directly from this split, and they are attainable if and only if the sign-bias vector belongs to the even-parity polytope for the upper bound or the odd-parity polytope for the lower bound. These conditions are necessary and sufficient under a mild regularity assumption that permits explicit construction of the achieving couplings via measurable selections. The work also supplies explicit trivariate copulas and a recursive method to build higher-dimensional extremal couplings, including non-symmetric ones for identical marginals.

Core claim

We study the sharp bounds of E[X1⋯Xd] when the univariate marginal distributions are known, but the dependence structure between them is unspecified. We propose a decomposition of the problem into a magnitude part and a sign part, and show that universal upper and lower bounds for the product expectation follow from the comonotonic coupling of the absolute values and properly chosen sign vectors. Under a mild regularity assumption, we give necessary and sufficient conditions for these universal bounds to be attainable. For the upper bound, the marginal sign-bias vector must belong to the even-parity polytope, while for the lower, the corresponding condition involves the odd-parity polytope.

What carries the argument

Decomposition into magnitude and sign components, where the sign-bias vector must belong to the even-parity polytope (upper bound) or odd-parity polytope (lower bound) for the comonotonic magnitude bounds to be attainable.

Load-bearing premise

The mild regularity assumption must hold so that measurable selections on the parity polytope exist and the extremal couplings can be constructed from the sign-bias vector.

What would settle it

A concrete set of marginal distributions whose sign-bias vector is in the even-parity polytope yet no joint distribution reaches the proposed upper bound would show the condition is not sufficient.

Figures

Figures reproduced from arXiv: 2604.22667 by Christopher Blier-Wong, Jinghui Chen.

Figure 1
Figure 1. Figure 1: Parity polytopes in dimension d = 3, shown inside the unit cube [0, 1]3 . The even-parity polytope P + 3 = conv{(1, 1, 1),(1, 0, 0),(0, 1, 0),(0, 0, 1)} (left, blue) and the odd-parity polytope P − 3 = conv{(1, 1, 0),(1, 0, 1),(0, 1, 1),(0, 0, 0)} (center, red) are congruent regular tetrahedra whose eight vertices together are exactly {0, 1} 3 . In the left and center panels, each tetrahedron vertex is lab… view at source ↗
Figure 2
Figure 2. Figure 2: Linear densities with (θ1, θ2, θ3) = (0.4, 0.2, −0.3). (a) Feasibility region: the cross-polytope |θ1| + |θ2| + |θ3| ≤ 1; the red dot represents the specific numerical example. (b) Support of the maximizing copula in [0, 1]3 ; each colour corresponds to one even-parity sign pattern. (c) Support of the minimizing copula; each colour corresponds to one odd-parity sign pattern. In (b) and (c), the black dot m… view at source ↗
Figure 3
Figure 3. Figure 3: Support of the maximizing copula for shifted exponential marginals ( view at source ↗
Figure 4
Figure 4. Figure 4: Support of the maximizing copula for heterogeneous shifted exponential marginals with view at source ↗
Figure 5
Figure 5. Figure 5: Support of the maximizing copula for N(µ, 1) marginals (µ = Φ−1 (0.75) ≈ 0.674). Each colour corresponds to one even-parity sign pattern; the black dot marks the junction (u0, u0, u0) = (0.25, 0.25, 0.25). Proof. By Proposition 2 with d = 3, the upper bound is attainable iff p(u) ≥ 1/3 for a.e. u. When µ ≥ 0, the logistic form (31) gives p(u) ≥ 1/2 ≥ 1/3 for all u ∈ (0, 1), since µG−1 (u)/σ2 ≥ 0. Theorem 1… view at source ↗
Figure 6
Figure 6. Figure 6: Support of the minimizing copula for N(µ, 1) marginals (µ = Φ−1 (0.25) ≈ −0.674). Each colour corresponds to one odd-parity sign pattern; the black dot marks the junction (u0, u0, u0) = (0.75, 0.75, 0.75). For the shifted exponential marginals of Appendices A.1 and A.2, Corollary 1 rules out the universal lower bound directly. For normal marginals with µ ̸= 0, unbounded support does not suffice to attain b… view at source ↗
read the original abstract

We study the sharp bounds of $\mathbb{E}[X_1\cdots X_d]$ when the univariate marginal distributions are known, but the dependence structure between them is unspecified. Maximizing products over non-negative variables is straightforward via the comonotonic coupling, but the problem is more subtle when the marginals can take both positive and negative values. Specifically, two negative realizations can be matched to yield a positive product, whereas a single negative realization necessarily yields a negative product. We propose a decomposition of the problem into a magnitude part and a sign part, and show that universal upper and lower bounds for the product expectation follow from the comonotonic coupling of the absolute values and properly chosen sign vectors. Under a mild regularity assumption, we give necessary and sufficient conditions for these universal bounds to be attainable. For the upper bound, the marginal sign-bias vector must belong to the even-parity polytope, while for the lower, the corresponding condition involves the odd-parity polytope. We construct the extremal couplings via measurable selections on the parity polytope whenever these conditions hold. We study the case of identical marginals in more detail and provide examples of non-symmetric extremal coupling that achieve the universal bounds. We explicitly construct the extremal copulas in three dimensions, and use a recursive parity decomposition to obtain higher-dimensional extremal copulas from the trivariate ones.

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 derives sharp upper and lower bounds on E[X1⋯Xd] given only the univariate marginals of the Xi (which may change sign). It decomposes the product into a comonotonic coupling of the absolute values |Xi| together with a sign vector whose marginal bias vector must lie in the even-parity polytope (upper bound) or odd-parity polytope (lower bound). Under a mild regularity assumption, these conditions are necessary and sufficient for attainability; the extremal joints are constructed by measurable selection on the parity polytope. Explicit trivariate copulas and a recursive construction for higher dimensions are supplied, together with examples for identical marginals.

Significance. If the claims hold, the work supplies the first complete N&S characterization of attainable bounds for signed products under marginal constraints, together with constructive copulas. This strengthens the literature on Fréchet–Hoeffding-type bounds and comonotonicity by handling sign changes rigorously, with potential applications to moment bounds in risk analysis and stochastic optimization.

major comments (2)
  1. [Abstract / construction of extremal couplings] Abstract and construction section: the sufficiency direction (attainability whenever the sign-bias vector lies in the appropriate parity polytope) rests entirely on the existence of a measurable selection under the invoked mild regularity assumption. The skeptic note correctly identifies that this step is load-bearing; if the assumption fails for discrete or singular marginals with mixed signs, the stated bounds are not always sharp even when the polytope condition holds. The manuscript must either (i) state the regularity assumption explicitly (e.g., as a condition on the marginal cdfs or supports) or (ii) supply a counter-example showing when selection fails, together with the resulting gap between the universal bound and the true sharp bound.
  2. [Recursive construction] Recursive parity decomposition (higher-d section): the reduction from d to d−1 via trivariate extremal copulas assumes that the sign-bias vector of the reduced problem remains inside the appropriate parity polytope after each marginalization step. No explicit verification is given that the polytope membership is preserved under the recursion; a counter-example or inductive argument is needed to confirm that the constructed coupling indeed attains the claimed universal bound in dimension d>3.
minor comments (2)
  1. Notation for the sign-bias vector and the even/odd parity polytopes should be introduced with a short self-contained definition (perhaps in a preliminary section) rather than only in the abstract.
  2. The three-dimensional explicit copula constructions would benefit from a small numerical table comparing the attained product expectation against the universal bound for a few mixed-sign marginals.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading, the positive overall assessment, and the constructive major comments. We address each point below and will revise the manuscript to incorporate explicit statements and additional arguments as requested.

read point-by-point responses

  1. Referee: [Abstract / construction of extremal couplings] Abstract and construction section: the sufficiency direction (attainability whenever the sign-bias vector lies in the appropriate parity polytope) rests entirely on the existence of a measurable selection under the invoked mild regularity assumption. The skeptic note correctly identifies that this step is load-bearing; if the assumption fails for discrete or singular marginals with mixed signs, the stated bounds are not always sharp even when the polytope condition holds. The manuscript must either (i) state the regularity assumption explicitly (e.g., as a condition on the marginal cdfs or supports) or (ii) supply a counter-example showing when selection fails, together with the resulting gap between the universal bound and the true sharp bound.

    Authors: We agree that the sufficiency claim relies on measurable selection and that the regularity assumption should be stated explicitly rather than left implicit. In the revised manuscript we will add a precise formulation: the marginal distributions are assumed to be atomless (i.e., their cumulative distribution functions are continuous). Under this condition the required measurable selection from the parity polytope exists by standard measurable-selection theorems, rendering the stated bounds sharp and attainable. We will also include a brief remark noting that the necessity part of the characterization holds without the continuity assumption, while sufficiency may fail for purely atomic marginals; in such cases the polytope condition remains necessary but the universal bound may not be attained. This revision directly implements option (i) requested by the referee. revision: yes

  2. Referee: [Recursive construction] Recursive parity decomposition (higher-d section): the reduction from d to d−1 via trivariate extremal copulas assumes that the sign-bias vector of the reduced problem remains inside the appropriate parity polytope after each marginalization step. No explicit verification is given that the polytope membership is preserved under the recursion; a counter-example or inductive argument is needed to confirm that the constructed coupling indeed attains the claimed universal bound in dimension d>3.

    Authors: We acknowledge that an explicit verification of polytope preservation under the recursion was omitted. In the revised version we will insert a short inductive argument in the higher-dimensional section. The base case (d=3) is already verified by the explicit trivariate copula construction. For the inductive step, we show that if the d-dimensional sign-bias vector lies in the even- (respectively odd-) parity polytope, then the (d−1)-dimensional marginal sign-bias vector obtained after integrating out one coordinate via the trivariate extremal copula also lies in the corresponding (d−1)-dimensional parity polytope. The argument uses the facet description of the parity polytopes and the fact that the trivariate coupling preserves the required parity relations among the sign probabilities. This establishes that the recursively constructed coupling attains the claimed universal bound for all d. revision: yes

Circularity Check

0 steps flagged

No circularity: bounds derived from comonotonic decomposition and polytope conditions

full rationale

The derivation decomposes E[product] into comonotonic absolute values plus sign vectors whose bias must lie in even- or odd-parity polytopes. Necessary and sufficient attainability conditions are stated under an explicit mild regularity assumption that guarantees measurable selections from the polytope. This is a direct existence argument using standard measure-theoretic tools, not a reduction of the claimed bounds to the inputs by construction. No fitted parameters are renamed as predictions, no self-citations are load-bearing for the core claim, and no ansatz or renaming of known results occurs. The paper is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 3 invented entities

The paper relies on standard results from coupling theory and convex geometry while introducing new structures (parity polytopes) whose properties are derived within the work; no free parameters are fitted to data.

axioms (2)
  • standard math Existence of comonotonic couplings for the absolute values of the random variables
    Standard fact from probability theory used to maximize the magnitude component.
  • domain assumption Existence of measurable selections from the parity polytopes under the mild regularity assumption
    Invoked to guarantee that extremal sign couplings can be constructed when the sign-bias vector lies in the polytope.
invented entities (3)
  • even-parity polytope no independent evidence
    purpose: Characterizes the set of attainable sign-bias vectors for the upper bound on the product expectation
    New convex set introduced to encode the parity constraints on sign combinations that produce positive products.
  • odd-parity polytope no independent evidence
    purpose: Characterizes the set of attainable sign-bias vectors for the lower bound on the product expectation
    Analogous new convex set for sign combinations that produce negative products.
  • sign-bias vector no independent evidence
    purpose: Encodes the marginal probabilities that each variable is positive or negative
    Derived directly from the given univariate marginal distributions.

pith-pipeline@v0.9.0 · 5541 in / 1678 out tokens · 63431 ms · 2026-05-08T09:13:24.252109+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

3 extracted references · 1 canonical work pages

  1. [1]

    Aliprantis, C. D. and Border, K. C. (2006).Infinite Dimensional Analysis: A Hitchhiker’s Guide. Springer, Berlin, 3 edition. Barvinok, A. (2002).A Course in Convexity, volume 54 ofGraduate Studies in Mathematics. American Mathematical Society. Bernard, C., Chen, J., R¨ uschendorf, L., and Vanduffel, S. (2023). Coskewness under dependence uncertainty.Stati...

    work page arXiv 2006

  2. [2]

    and R¨ uschendorf, L

    Puccetti, G. and R¨ uschendorf, L. (2013). Sharp bounds for sums of dependent risks.Journal of Applied Probability, 50(1):42–53. Puccetti, G. and R¨ uschendorf, L. (2015). Computation of sharp bounds on the expected value of a supermodular function of risks with given marginals.Communications in Statistics - Simulation and Computation, 44(3):705–718. R¨ u...

    2013

  3. [3]

    On the positive branch, rµ(u) = (1 −p (w))/p(w) ≤ 1 since p(u) ≥ 1/2, so s+(u) = 1 −r µ(u)/2 ∈ [1/2, 1)

    Theorem 1 then guarantees that the specialization of (18) to identical marginals produces the correct marginals and attains the bound; the mixing function s+ in (32) is obtained by matching the even-parity weights (16) as in Section 5.1. On the positive branch, rµ(u) = (1 −p (w))/p(w) ≤ 1 since p(u) ≥ 1/2, so s+(u) = 1 −r µ(u)/2 ∈ [1/2, 1). Similarly, for...

    2023