Stochastic Ordering of Dependent Systems under Transformation Models and Archimedean Copulas

Idir Arab; Milto Hadjikyriakou; Paulo Eduardo Oliveira

arxiv: 2604.26026 · v1 · submitted 2026-04-28 · 🧮 math.PR

Stochastic Ordering of Dependent Systems under Transformation Models and Archimedean Copulas

Idir Arab , Milto Hadjikyriakou , Paulo Eduardo Oliveira This is my paper

Pith reviewed 2026-05-07 14:53 UTC · model grok-4.3

classification 🧮 math.PR

keywords stochastic orderingArchimedean copulassystem reliabilitytransformation modelsstochastic dominanceparallel systemsseries systemsdependent components

0 comments

The pith

Stochastic dominance conditions for dependent system lifetimes follow from transformation monotonicity and Archimedean copula generator properties.

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

The paper derives conditions under which the lifetime of one system stochastically dominates another for parallel, series, and (n-k)-out-of-n structures. Component lifetimes are dependent through an Archimedean copula and arise from monotone transformations applied to a shared baseline distribution. The conditions use super-additivity or Schur-type properties of the copula generator together with the direction of the transformation monotonicity. A sympathetic reader would care because the criteria remain tractable even with dependence and heterogeneity, extending comparisons that were previously limited to independent components and clarifying how both factors jointly shape overall reliability.

Core claim

For parallel, series and (n-k)-out-of-n systems whose components have marginal distributions obtained by transforming a common baseline, stochastic dominance of system lifetimes holds when the transformation is monotone in the appropriate direction and the generator of the Archimedean copula satisfies super-additivity or Schur-type convexity. This framework unifies several transformation models including proportional hazard, proportional reversed hazard, and proportional odds, and supplies explicit dominance rules that incorporate the strength and direction of dependence.

What carries the argument

Archimedean copulas whose generators obey super-additivity or Schur-type conditions, paired with monotone transformations of a baseline distribution, to produce stochastic ordering of the resulting system lifetime distributions.

If this is right

Under the stated generator conditions, increasing parameter heterogeneity in a given direction produces stochastic dominance for parallel systems.
The same generator conditions yield dominance in the opposite direction for series systems.
The ordering results continue to hold for (n-k)-out-of-n systems under the corresponding structural adjustments.
The criteria remain valid when dependence is present and thereby extend all earlier comparisons that assumed independence.

Where Pith is reading between the lines

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

The ordering rules could be applied directly to compare alternative system designs by checking only the generator property and the transformation direction rather than simulating full joint distributions.
The same super-additivity arguments might be adapted to bound reliability under other parametric dependence families that admit analogous generator representations.
In practice the conditions supply sensitivity statements showing how strengthening dependence in one direction can offset or reinforce the effect of parameter heterogeneity.

Load-bearing premise

The joint distribution of component lifetimes is exactly given by an Archimedean copula applied to the transformed marginals.

What would settle it

Explicit computation of the survival function for a concrete parallel system with a super-additive generator such as the Clayton family and a monotone transformation that satisfies the stated monotonicity would falsify the claim if the predicted dominance inequality fails to hold.

Figures

Figures reproduced from arXiv: 2604.26026 by Idir Arab, Milto Hadjikyriakou, Paulo Eduardo Oliveira.

**Figure 1.** Figure 1: Plots for (αℓ − αm) ∂F(α) 2:3 ∂αℓ − ∂F(α) 2:3 ∂αm : (ℓ, m) = (1, 2) (solid), (ℓ, m) = (1, 3) (dashed), (ℓ, m) = (2, 3) (dotted). sign, the case (ℓ, m) = (1, 2) changes sign as x ranges (0, +∞). Hence, the map α 7−→ F (α) 2:3 (x) is neither Schur-convex nor Schur-concave. 4 Stochastic Comparisons for specific transformation models We now demonstrate how the proposed framework can be applied to standard … view at source ↗

read the original abstract

We study stochastic ordering of system lifetimes with dependent and heterogeneous components whose marginal distributions are obtained through transformations of a common baseline. The dependence structure is modeled via Archimedean copulas, allowing for a unified treatment of several transformation-based models, including proportional hazard, proportional reversed hazard rate and proportional odds families. For parallel, series and $(n-k)$-out-of-$n$ systems, we derive conditions for stochastic dominance based on monotonicity of the transformation and structural properties of the copula generators, formulated through super-additivity and Schur-type arguments. The results provide tractable criteria that extend existing comparisons beyond independence and illustrate the combined effect of dependence and parameter heterogeneity on system reliability.

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 extends stochastic ordering to dependent reliability systems by combining Archimedean copulas with transformation models and gives explicit dominance conditions for the usual structures.

read the letter

The core contribution is a set of conditions under which one system lifetime stochastically dominates another when the components are heterogeneous, follow a transformation model, and are linked by an Archimedean copula. They handle parallel, series, and (n-k)-out-of-n systems in one framework that covers proportional hazards, reversed hazards, and odds models. The arguments rely on monotonicity of the transformation functions together with super-additivity and Schur-convexity properties of the copula generator. This moves the comparison past the independent case without requiring new machinery beyond what is already standard in the literature on copula-based orders. The setup is clean and the criteria look usable once the generator and the baseline are fixed. That is the useful part: it supplies tractable rules that show how dependence strength and parameter spread affect the order. The derivations stay within established properties of Archimedean generators and do not introduce hidden restrictions on the baseline distributions. The paper therefore earns credit for making the extension explicit and for keeping the conditions checkable from the generator and the transformation. The main limitation is that the results remain purely theoretical. There is no numerical illustration of how often the conditions hold for typical parameter ranges, nor any discussion of how sensitive the dominance is to small violations of super-additivity. Readers will still need to verify the generator properties case by case. The work is aimed at people already working in stochastic orders for reliability or survival analysis. Someone who knows the independent-case results will see the extension immediately and can judge whether the new conditions are worth applying in their setting. It is narrow enough that a general probability audience will not find it essential, but it is solid enough for the subfield. I would send it to a serious referee. The claims are specific, the tools are standard, and the gaps are the usual ones for this style of paper rather than load-bearing flaws.

Referee Report

2 major / 2 minor

Summary. The manuscript studies stochastic ordering of lifetimes for parallel, series, and (n-k)-out-of-n systems whose components are heterogeneous and dependent. Marginal distributions arise from monotonic transformations of a common baseline (covering proportional hazards, proportional reversed hazards, and proportional odds families), while dependence is captured by Archimedean copulas. Conditions for stochastic dominance are derived from monotonicity of the transformation functions together with super-additivity and Schur-convexity properties of the copula generators.

Significance. If the central derivations are correct, the paper supplies a unified, tractable set of criteria that extend existing stochastic-order results beyond the independent case. The approach integrates standard tools from reliability theory and copula theory, and the explicit use of generator properties yields concrete, checkable conditions that illustrate the joint effect of dependence strength and parameter heterogeneity on system reliability.

major comments (2)

[§3.2, Theorem 3.3] §3.2, Theorem 3.3: the proof that the joint survival function is Schur-convex under the stated generator conditions invokes super-additivity of the inverse generator; however, the argument does not explicitly verify that the super-additivity inequality remains strict when the transformation parameters differ across components, which is required for the claimed strict stochastic ordering in the heterogeneous case.
[§4.1, Eq. (4.2)] §4.1, Eq. (4.2): the reduction of the (n-k)-out-of-n survival probability to a sum over order statistics relies on the convexity of the generator; the manuscript does not address the boundary case in which the generator is linear (independence limit), where the claimed dominance collapses to equality and the result becomes vacuous.

minor comments (2)

[§2] The notation for the baseline cumulative hazard and the transformation function φ is introduced without a dedicated preliminary subsection; a short table collecting the three families (PH, PRH, PO) and their corresponding φ would improve readability.
[§5] Several citations to the independent-case literature appear only in the introduction; adding one or two explicit comparisons in the discussion of the new dependence-adjusted bounds would clarify the precise extension achieved.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading, positive assessment, and recommendation for minor revision. The comments highlight opportunities to strengthen the explicitness of our arguments. We address each major comment point by point below and indicate the revisions we will make to the manuscript.

read point-by-point responses

Referee: §3.2, Theorem 3.3: the proof that the joint survival function is Schur-convex under the stated generator conditions invokes super-additivity of the inverse generator; however, the argument does not explicitly verify that the super-additivity inequality remains strict when the transformation parameters differ across components, which is required for the claimed strict stochastic ordering in the heterogeneous case.

Authors: We thank the referee for this observation. The super-additivity of the inverse generator is strict whenever the generator is strictly convex (as assumed in the theorem) and the transformation parameters differ, because the marginal transformations are strictly increasing. To make this verification explicit, we will add a clarifying sentence in the proof of Theorem 3.3 stating that the inequality is strict for distinct parameters. This is a minor addition that directly addresses the concern for strict stochastic ordering in the heterogeneous case. revision: yes
Referee: §4.1, Eq. (4.2): the reduction of the (n-k)-out-of-n survival probability to a sum over order statistics relies on the convexity of the generator; the manuscript does not address the boundary case in which the generator is linear (independence limit), where the claimed dominance collapses to equality and the result becomes vacuous.

Authors: We agree that the linear generator corresponds to the independence limit, in which the claimed stochastic dominance reduces to equality. This is a natural boundary case where the dependence structure vanishes. We will add a brief remark in Section 4.1 noting that the result is non-vacuous for strictly convex generators (capturing positive dependence), while the linear case recovers the corresponding ordering under independence. This clarifies the scope of the theorem without altering its statements. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The derivations for stochastic dominance in parallel, series, and (n-k)-out-of-n systems apply standard properties of Archimedean copula generators (decreasing, convex, super-additive inverse) together with monotonicity of transformation functions and Schur-convexity arguments to the joint survival functions under PH, PRH, and PO families. These steps rest on externally established copula theory and reliability ordering results rather than any self-definition, fitted parameter renamed as prediction, or load-bearing self-citation chain. The central claims therefore remain independent of the paper's own inputs and do not reduce by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 0 invented entities

The paper rests on standard domain assumptions from copula theory and reliability modeling without introducing new free parameters or postulated entities; all elements appear drawn from established literature.

axioms (3)

domain assumption Archimedean copulas can model the dependence structure among component lifetimes
Explicitly stated as the modeling choice for dependence in the abstract.
domain assumption Marginal distributions arise from monotonic transformations of a common baseline
Core modeling assumption enabling the unified treatment of proportional hazard, reversed hazard, and odds families.
standard math Super-additivity and Schur-type properties of copula generators imply stochastic ordering under the stated conditions
Invoked as the basis for the derived dominance conditions.

pith-pipeline@v0.9.0 · 5416 in / 1485 out tokens · 79834 ms · 2026-05-07T14:53:32.690072+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

[1]

doi: 10.1016/j.orl.2020.12.009. 22 M. Shaked and J. G. Shanthikumar.Stochastic orders. Springer Series in Statistics, 2007. A. Sklar. Fonctions de r´ epartition andimensions et leurs marges.Ann. I.S.U.P., 54(1-2):3–6,

work page doi:10.1016/j.orl.2020.12.009 2020
[2]

With an introduction by Denis Bosq. P. Zhao, Y. Zhang, and J. Qiao. On extreme order statistics from heterogeneous Weibull variables.Statistics, 50(6):1376–1386, 2016. doi: 10.1080/02331888.2016.1230859. 23

work page doi:10.1080/02331888.2016.1230859 2016

[1] [1]

doi: 10.1016/j.orl.2020.12.009. 22 M. Shaked and J. G. Shanthikumar.Stochastic orders. Springer Series in Statistics, 2007. A. Sklar. Fonctions de r´ epartition andimensions et leurs marges.Ann. I.S.U.P., 54(1-2):3–6,

work page doi:10.1016/j.orl.2020.12.009 2020

[2] [2]

With an introduction by Denis Bosq. P. Zhao, Y. Zhang, and J. Qiao. On extreme order statistics from heterogeneous Weibull variables.Statistics, 50(6):1376–1386, 2016. doi: 10.1080/02331888.2016.1230859. 23

work page doi:10.1080/02331888.2016.1230859 2016