Gibbs conditioning principle for log-concave independent random variables

Eric Cator; Pablo A. Ferrari

arxiv: 2512.24910 · v2 · submitted 2025-12-31 · 🧮 math.PR

Gibbs conditioning principle for log-concave independent random variables

Eric Cator , Pablo A. Ferrari This is my paper

Pith reviewed 2026-05-16 18:12 UTC · model grok-4.3

classification 🧮 math.PR

keywords gibbs conditioning principlelog-concave distributionstilted measuresstochastic orderingefron theoremconditional convergencenonnegative integerslarge deviations

0 comments

The pith

Log-concave independent random variables satisfy the Gibbs conditioning principle when condensation is prevented.

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

This paper proves that for a sequence of independent random variables with log-concave laws on the nonnegative integers, the conditional distribution given that their sum exceeds a specific threshold converges weakly to an exponentially tilted version of the original sequence. The tilt parameter is chosen strictly between 1 and the supremum value that keeps the normalizing constants finite. A sympathetic reader cares because the result describes the typical configuration under a rare large-sum event without fixing the exact sum value, extending classical conditioning results to a wide family of discrete distributions. The argument proceeds by invoking Efron's theorem, which supplies stochastic ordering of the canonical conditional measures for log-concave laws, and then transferring that ordering to the tilted versions.

Core claim

The paper proves that the Gibbs conditioning principle holds: for log-concave probabilities ν_i on the nonnegative integers with λ^max greater than 1, and for any fixed λ* in (1, λ^max), the law of the sequence X conditioned on the event that the sum S_n exceeds R*_n converges weakly to the law of the tilted sequence X^{λ*} as n tends to infinity, provided a technical condition prevents condensation. The proof rests on the fact that log-concavity implies, via Efron's theorem, that the canonical measures are stochastically ordered in the conditioning value k; this ordering lifts to the family of conditioned tilted measures and yields the desired limit.

What carries the argument

Log-concavity of each probability mass function ν_i, expressed by the inequality ν_i(x+1) ν_i(x-1) ≤ ν_i(x)^2, which activates Efron's theorem to produce stochastic ordering of the canonical conditional distributions with respect to the sum value.

If this is right

The conditional distribution converges in the weak topology to the infinite product of the tilted marginals ν_i^{λ*}.
The same convergence holds uniformly over sequences of log-concave laws that obey the non-condensation requirement.
The family of conditioned tilted measures remains stochastically ordered with respect to the tilt parameter λ.
The result applies directly to any concrete family of log-concave distributions on the nonnegative integers for which the normalizing constants remain finite in a neighborhood of 1.

Where Pith is reading between the lines

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

The ordering technique may extend to other classes of discrete distributions if an analogous stochastic monotonicity can be established by different means.
The limit object supplies a practical way to sample from the conditional law for large but finite n by simply drawing from the tilted product measure.
The non-condensation condition might be characterized more explicitly in terms of the decay rates of the tail probabilities of the tilted variables.

Load-bearing premise

A technical condition must hold to prevent condensation in the tilted measures, together with the existence of λ^max strictly larger than 1 and the choice of λ* inside (1, λ^max).

What would settle it

A concrete sequence of log-concave distributions ν_i satisfying λ^max >1 and the non-condensation condition for which the conditioned law P(X ∈ · | S_n > R*_n) fails to converge to the law of X^{λ*}.

read the original abstract

Let $\nu_1,\nu_2,\dots$ be a sequence of probabilities on the nonnegative integers, and $X=(X_1,X_2, \dots)$ be a sequence of independent random variables $X_i$ with law $\nu_i$. For $\lambda>0$ denote $Z^\lambda_i:= \sum_x \lambda^x\nu_i(x)$ and $\lambda^{\max}:= \sup\{\lambda>0: Z^\lambda_i<\infty \text{ for all }i\}$, and assume $\lambda^{\max}>1$. For $\lambda<\lambda^{\max}$, define the tilted probability $\nu_i^{\lambda}(x):= \lambda^x\nu_i(x)/Z^{\lambda}_i$, and let $X^\lambda$ be a sequence of independent variables $X^\lambda_i$ with law $\nu^{\lambda}_i$, and denote $S^\lambda_n:= X^{\lambda}_1+\dots+X^{\lambda}_n$, with $S_n=S^1_n$. Choose $\lambda^*\in(1,\lambda^{\max})$ and denote $R^*_n:= E (S^{\lambda^*}_n)$. The Gibbs Conditioning Principle (GCP) holds if $P(X\in\cdot|S_n>R^*_n)$ converges weakly to the law of $X^{\lambda^*}$, as $n\to\infty$. We prove the GCP for log-concave $\nu_i$'s, meaning $\nu_i(x+1)\,\nu_i(x-1) \le ( \nu_i(x))^2$, subject to a technical condition that prevents condensation. The canonical measures are the distributions of the first $n$ variables, conditioned on their sum being $k$. Efron's theorem states that for log-concave $\nu_i$'s, the canonical measures are stochastically ordered with respect to $k$. This, in turn, leads to the ordering of the conditioned tilted measures $P(X^\lambda\in\cdot|S^\lambda_n>R^*_n)$ in terms of $\lambda$. This ordering is a fundamental component of our proof.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note private letter to a colleague

The paper proves the Gibbs conditioning principle for log-concave discrete distributions on nonnegative integers by using Efron's theorem for stochastic ordering of canonical measures.

read the letter

The main result is a proof that the Gibbs conditioning principle holds for independent log-concave random variables on the nonnegative integers. Under the assumptions that lambda max exceeds 1 and lambda star sits in (1, lambda max), the conditional law given the sum exceeds its tilted mean converges weakly to the exponentially tilted law, provided a no-condensation condition is met. The argument relies on Efron's theorem to obtain stochastic ordering of the canonical measures with respect to the conditioning value, which then orders the conditioned tilted measures and delivers the limit. This is a genuine extension because it moves beyond the usual Poisson or geometric cases to the full log-concave class, which covers many standard discrete distributions. The strategy is direct and stays inside tools already in the literature, so the connection between log-concavity, ordering, and the conditioning limit is easy to follow on the surface. The soft spot is the unspecified technical condition that prevents condensation. The abstract gives no indication of how restrictive it is or how often it holds for typical sequences, which leaves the practical reach of the theorem unclear. Since only the abstract is available, the actual derivation steps and any edge cases in applying Efron's theorem cannot be checked. The assumptions on lambda are standard and do not appear circular. This work is aimed at probabilists who study conditional limit theorems and large deviations for sums of independent discrete random variables. A reader already comfortable with the classical GCP and Efron's theorem will see the value in the extension. It deserves a serious referee because the claim is specific, the method is appropriate, and the gaps are technical rather than foundational. I would send it out for peer review rather than desk reject.

Referee Report

2 major / 0 minor

Summary. The manuscript claims to prove the Gibbs Conditioning Principle (GCP) for independent log-concave random variables X_i on the nonnegative integers. Assuming λ^max > 1 and choosing λ* ∈ (1, λ^max), it establishes that P(X ∈ · | S_n > R_n*) converges weakly to the law of the tilted variables X^{λ*}, where R_n* = E[S_n^{λ*}]. The argument relies on Efron's theorem to obtain stochastic ordering of the canonical measures for log-concave ν_i, which induces ordering on the conditioned tilted laws P(X^λ ∈ · | S_n^λ > R_n*), subject to an unspecified technical condition preventing condensation.

Significance. If the result holds, this work extends the GCP to the broad class of log-concave discrete distributions on nonnegative integers, including Poisson and binomial laws. The approach via Efron's theorem for stochastic ordering of canonical measures offers a clean, potentially reusable technique that avoids direct large-deviation analysis. This could strengthen connections between conditioning principles and combinatorial probability.

major comments (2)

Abstract: The technical condition that prevents condensation is left unspecified. This condition is load-bearing, as it is required for the tilted measures to be well-defined and for the weak convergence to hold under the stated assumptions on λ^max and λ*.
Abstract: The manuscript states that Efron's theorem yields stochastic ordering of the canonical measures, which in turn orders the conditioned tilted measures P(X^λ ∈ · | S_n^λ > R_n^*). No derivation or verification of this implication is supplied, leaving open whether the ordering step is rigorous or contains gaps.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments. We address the major comments point by point below.

read point-by-point responses

Referee: Abstract: The technical condition that prevents condensation is left unspecified. This condition is load-bearing, as it is required for the tilted measures to be well-defined and for the weak convergence to hold under the stated assumptions on λ^max and λ*.

Authors: We agree that the no-condensation condition is central and should be stated explicitly. In the revised version we will update the abstract to include a concise description of this condition (ensuring linear growth of the variance of the tilted sums and preventing mass concentration that would invalidate the limiting law). revision: yes
Referee: Abstract: The manuscript states that Efron's theorem yields stochastic ordering of the canonical measures, which in turn orders the conditioned tilted measures P(X^λ ∈ · | S_n^λ > R_n^*). No derivation or verification of this implication is supplied, leaving open whether the ordering step is rigorous or contains gaps.

Authors: The abstract is a summary; the body derives the stochastic ordering directly from Efron's theorem applied to log-concave ν_i and shows how the ordering on canonical measures transfers to the conditioned tilted laws via monotonicity in the tilting parameter λ. We will add a brief clarifying sentence in the abstract and ensure the relevant proof section highlights the steps explicitly. revision: partial

Circularity Check

0 steps flagged

No significant circularity; proof relies on external Efron's theorem

full rationale

The abstract presents a direct proof of the GCP under log-concavity (explicitly defined as ν_i(x+1)ν_i(x-1) ≤ (ν_i(x))^2) plus a technical no-condensation condition. It invokes Efron's theorem for stochastic ordering of canonical measures, an independent external result, which then yields ordering of the conditioned tilted laws. The parameters λ^max > 1 and λ* ∈ (1, λ^max) are standard for the exponential tilt to be defined and for the mean R_n^* to grow; they are not fitted to the target limit. No self-citations, self-definitional steps, fitted inputs renamed as predictions, or ansatzes appear. The derivation chain is self-contained against external benchmarks and does not reduce to its inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters or invented entities. The proof rests on standard mathematical background plus Efron's theorem.

axioms (1)

standard math Efron's theorem on stochastic ordering of canonical measures for log-concave distributions
Invoked to obtain ordering of the conditioned tilted measures in terms of λ.

pith-pipeline@v0.9.0 · 5657 in / 1146 out tokens · 47533 ms · 2026-05-16T18:12:49.170219+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

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

IndisputableMonolith/Foundation/AbsoluteFloorClosure.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We prove the GCP for log-concave ν_i's, meaning ν_i(x+1) ν_i(x-1) ≤ (ν_i(x))^2, subject to a technical condition that prevents condensation. ... Efron's theorem states that for log-concave ν_i's, the canonical measures are stochastically ordered with respect to k.

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.