Pith. sign in

REVIEW 1 major objections 3 minor 36 references

Fractional extreme distributions

T0 review · 1 major / 3 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Fractional-order versions of the defining equations of Weibull, Fréchet, and Gumbel laws have unique solutions with explicit stable-subordinator representations.

desk verdict A new family of fractional extreme distributions with solid main theorems, undercut by a sign error in the definition of the progressive derivative that breaks Theorem 1.3 as written. read the letter →

arxiv 1908.00584 v1 pith:XOFVOH3U submitted 2019-08-01 math.PR math.CA

Lotfi Boudabsa , Thomas Simon , Pierre Vallois This is my paper · ORCID
classification math.PRmath.CA MSC 26A3333E1245E1060E0560E1560G52
keywords fractionaldifferentialequationsextremevaluedistributionsalpha-stablesubordinatorKilbas-SaigofunctionLeRoyMittag-Lefflercompletemonotonicityinfinitedivisibility
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper extends the three classical extreme-value distributions—Weibull, Fréchet, and Gumbel—to fractional order. For every order $\alpha\in[0,1]$ and positive parameters $\lambda,\rho$, the authors prove that the fractional differential equations obtained by replacing the ordinary derivative with a Liouville fractional derivative have exactly one distribution-function solution, and they identify the solving random variables as the classical extreme laws multiplied or shifted by functionals of an independent $\alpha$-stable subordinator. The power-law examples are solved by Kilbas-Saigo functions in the Weibull and Fréchet cases and by Le Roy functions in the Gumbel case, and the stochastic representations yield new analytic properties of those special functions. The family traces a continuous arc from Pareto-type and logistic laws at $\alpha=0$ to the classical extreme families at $\alpha=1$.

What carries the argument

The load-bearing object is the infinite $\beta$ product $T(a,b,c)=\prod_{n\ge0}\frac{a+nb+c}{a+nb}B_{a+nb,c}$, together with the known identities that identify it, up to Gamma factors, with integral functionals of the $\alpha$-stable subordinator; for example $\int_0^\infty((1-\sigma_t^{(\alpha)})_+)^{\rho-\alpha}dt$ has the law of $\frac{\Gamma(\rho+1-\alpha)}{\Gamma(\rho+1)}T(1,\rho^{-1},(1-\alpha)\rho^{-1})$. These identities convert the formal series solutions of the fractional equations into explicit multiplicative or additive stochastic representations. The analytic counterparts are the Kilbas-Saigo functions $E_{\alpha,m,l}$, three-parameter Mittag-Leffler-type series, and the Le Roy functions $L_\alpha(z)=\sum_{n\ge0}z^n/(n!)^\alpha$; moment determinacy via a classical convergence criterion and Mellin transforms carry the analytic consequences.

What would settle it

Simulate an $\alpha$-stable subordinator for a fixed $\alpha\in(0,1)$ and $\rho>0$, compute the integral $\int_0^\infty((1-\sigma_t^{(\alpha)})_+)^{\rho-\alpha}dt$, and check whether the resulting survival function matches the series $E_{\alpha,\rho/\alpha,\rho/\alpha-1}(-\lambda x^\rho)$ at several values of $x$; a mismatch at any single point would falsify the Weibull-type representation. Equivalently, test the Mellin transform identity $E[W_{\alpha,\lambda,\rho}^s]=(\rho\alpha/\lambda)^{s/\rho}\Gamma(1+s/\rho)[\rho+(1-\alpha);\rho]_{-s}/[\rho;\rho]_{-s}$ for $s\in(-\rho,\rho)$, since the right-hand side must be analytic in that strip.

Watch

Extended reading notes

Core claim

The paper claims that each of the three classical extreme-value distributions can be deformed continuously through a fractional order $\alpha\in[0,1]$: replacing the ordinary derivative in their defining equations by a Liouville fractional derivative yields, for every $\lambda,\rho>0$, a unique distribution function. The solving laws are explicit. If $W_\rho$, $F_\rho$, and $G$ denote the usual Weibull, Fréchet, and Gumbel variables and $\sigma^{(\alpha)}$ an independent $\alpha$-stable subordinator, then $W_{\alpha,\lambda,\rho}=W_\rho(\lambda\int_0^\infty((1-\sigma_t^{(\alpha)})_+)^{\rho-\alpha}dt)^{-1/\rho}$, $F_{\alpha,\lambda,\rho}=F_\rho(\lambda\int_0^\infty(1+\sigma_t^{(\alpha)})^{-\rho-\alpha}dt)^{1/\rho}$, and $G_{\alpha,\lambda}=\lambda^{-1}(G-G_\alpha)$ with $G_\alpha=\log\int_0^\infty e^{-\sigma_t^{(\alpha)}}dt$. At $\alpha=1$ the equations reduce to the classical ones, while at $\alpha=0$ the solutions become Pareto-type and logistic laws, so the family traces an arc between those distributions. The same explicit laws are then used to prove analytic properties of the Kilbas-Saigo and Le Roy special functions.

Load-bearing premise

The construction rests on previously established equalities in law between certain infinite products of $\beta$-distributed random variables and integral functionals of an $\alpha$-stable subordinator; if those equalities fail for the parameter ranges used here ($\rho>0$, $\alpha\in(0,1)$), the explicit formulas for the fractional extreme laws and the analytic results built on them would not follow.

Editorial extensions

If this is right

  • For every $\alpha\in[0,1]$, $\lambda,\rho>0$, each of the three fractional equations has exactly one distribution-function solution, so the fractional-extreme family is a well-defined interpolation from Pareto and logistic laws at $\alpha=0$ to the Weibull, Fréchet, and Gumbel laws at $\alpha=1$.
  • The explicit representations yield Mellin transforms and exact tail asymptotics for the new laws, such as $f^W_{\alpha,\lambda,\rho}(x)\sim(\rho/(\lambda\Gamma(1-\alpha)))x^{-\rho-1}$ as $x\to\infty$ and $f^F_{\alpha,\lambda,\rho}(x)\sim C\,x^{\rho+\alpha-1}$ as $x\to0$ for an explicit constant $C$.
  • The Kilbas-Saigo function $E_{\alpha,m,m-1}(-x)$ is completely monotone for $\alpha\in(0,1]$ and $m>0$, with an explicit Bernstein representation; this settles a previously open question.
  • Optimal uniform hyperbolic bounds are obtained for the Kilbas-Saigo functions and for the generalized Mittag-Leffler functions $\Gamma(\beta)E_{\alpha,\beta}(-x)$, extending the classical two-sided bounds for $E_\alpha(-x)$.
  • The fractional extreme laws are infinitely divisible in certain parameter regimes: $W_{\alpha,\lambda,\rho}$ is a generalized Gamma convolution when $\rho\le1$, $W$ and $F$ are hyperbolically completely monotone when $\rho\le1-\alpha$, and $G_{\alpha,\lambda}$ belongs to the extended Thorin class for all $\alpha$.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: The single parameter $\alpha$ could serve as a tail-interpolation index for real data; one could estimate it from the slope of the log-survival function or from ratios of sample moments, a statistical use the paper does not develop.
  • Inference: The convex-order and peacock structure observed for the underlying beta products may extend to the fractional extreme variables themselves, implying monotonicity in $\alpha$ of quantiles or risk measures; this is not established in the paper.
  • Inference: The identity $W_{\alpha,1,\rho}\,F_{1-\alpha,\rho,\rho}^{-1}=(L_1L_2/L_3)^{1/\rho}$ with independent unit exponentials $L_i$ suggests a duality $\alpha\leftrightarrow1-\alpha$ in the family that could support symmetric estimation or model selection for the fractional order, a direction the paper leaves implicit.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

1 major / 3 minor

Summary. The paper introduces fractional analogues of the three classical extreme value distributions. For α ∈ [0,1], λ,ρ > 0, Theorems 1.1, 1.2, and 1.3 assert that the fractional differential equations (1.5), (1.7), and (1.9) each have a unique distribution function, and that the corresponding random variables admit explicit stochastic representations involving an independent α-stable subordinator and an exponential variable. The proofs combine uniqueness via alternating-series fixed-point arguments (Section 2), existence via Kilbas-Saigo and Le Roy functions and their moment determinacy (Section 3), and a substantial body of analytic consequences (Section 4): complete monotonicity, Mellin transforms and density asymptotics, optimal hyperbolic bounds, infinite divisibility properties, and new asymptotics for the Le Roy function. The appendix fixes the conventions for fractional integrals and derivatives and collects the needed properties of Barnes' double Gamma function.

Significance. If the main theorems are correct, the paper gives a coherent and elegant family of 'fractional extreme' laws interpolating between the classical Weibull, Fréchet, and Gumbel distributions and Pareto/logistic-type laws, with the stable subordinator playing a structurally natural role. The analytic spin-offs are also valuable: the complete monotonicity characterization of Kilbas-Saigo functions resolves a question raised in earlier literature, and the uniform hyperbolic bounds and Le Roy asymptotics are new and cleanly derived from stochastic representations. The paper's reliance on prior results from [25] for the beta-product identities is acceptable practice for a research article, though it means the stochastic representations are not self-contained. However, one load-bearing internal inconsistency in the definition of the progressive Liouville derivative on the line makes Theorem 1.3 ill-posed as written; this is a local sign error but it directly affects one of the three central existence theorems.

major comments (1)
  1. [Appendix A.1.3 and Theorem 1.3 (§3.3, Eq. (1.9))] Appendix A.1.3 defines the progressive Liouville derivative on R by D^α_+ f = −d/dx(I^{1−α}_+ f), with the boundary case D^1_+ described as the usual derivative. Under this definition, the claimed identity D^α_+F(x) = λ^α e^{λx}\bar F(x) in Theorem 1.3 is false. For \bar F(x)=L_α(−e^{λx}) = Σ_{n≥0}(−1)^n e^{λnx}/(n!)^α and F=1−\bar F, a direct computation gives I^{1−α}_+F = Σ_{n≥1}(−1)^{n+1}(λn)^{α−1}e^{λnx}/(n!)^α and hence d/dx I^{1−α}_+F = λ^α e^{λx}\bar F(x). The appendix definition therefore yields D^α_+F = −λ^α e^{λx}\bar F(x), contradicting Eq. (1.9). The proof of Theorem 1.3 is compatible only with the opposite sign convention, D^α_+ = +d/dx(I^{1−α}_+·), which is also the convention consistent with the classical α=1 case F'=e^x\bar F. This is an internal inconsistency in a central existence theorem, not merely a typo; it needs a local but substantive correction, either in the sign of the operator in A.1.3 or in the sign of the right-hand side of (1.9).
minor comments (3)
  1. [Propositions 4.7, 4.15, 4.17] Several secondary results are stated with proofs that omit essential details: Proposition 4.7 says 'We omit details' for the residue computation of the density asymptotics at zero, and Propositions 4.15 and 4.17 leave the single-intersection stochastic-order verification to the reader. These are not blocking for the central existence claims, but for publication the omitted arguments should be supplied or replaced by precise references.
  2. [Proposition 4.21] The displayed formula for log E[e^{sG_{α,λ}}] contains a sign typo: the second term should be (1−α) log Γ(1−sλ^{-1}) rather than (1−α) log Γ(1+sλ^{-1}); the Lévy measure expression written immediately afterward is consistent with the corrected version.
  3. [Appendix A.2, Eq. (4.20)] The phrase 'anxiolytic extension' in the sentence accompanying Eq. (4.20) is presumably a typo for 'analytic extension', since the intended meaning is that the right-hand side is understood analytically when z is a non-positive integer.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the central existence and uniqueness proofs are self-contained; the cited identities from [25] and [11] are prior parameter-free results, not restatements of the target equations.

full rationale

The paper's derivation chain is not circular in the sense defined here. Section 2 proves uniqueness for fractional hazard-rate equations directly, by converting each equation into a fixed-point problem for the operator A^{alpha,h} and then showing the alternating series converges; this does not presuppose the existence or form of the fractional extreme laws. Section 3 establishes existence by identifying the unique candidate series with Kilbas-Saigo functions and then showing, via Carleman's criterion, that these functions are Laplace transforms of positive random variables. The key external inputs are identities from [25] equating certain infinite beta products with integral functionals of an alpha-stable subordinator, and the Gumbel-case identity E[e^{nG_alpha}]=(n!)^{1-alpha} from [11]. Although [25] and [21] share an author with the present paper, these are published results with their own proofs and with assumptions that do not include the target statements (1.5), (1.7), or (1.9); they are not fitted parameters or renamed predictions, and the target laws are not used as hypotheses in those cited papers. The analytical results of Section 4 are consequences of the stochastic representations, not inputs to them. No data are fitted and no quantity is called a prediction after being used as an input. One non-circular concern should be flagged separately: Appendix A.1.3 defines the progressive Liouville derivative on the line with a minus sign, D^alpha_+ f = -d/dx(I^{1-alpha}_+ f), whereas the check in the proof of Theorem 1.3 appears to use the opposite sign; if sustained, this is a correctness defect in the Gumbel case, but it is not a circularity because it does not make the conclusion identical to an assumption.

Assumptions & free parameters 0 free parameters · 5 assumptions · 3 invented entities

No parameters are fitted to data; α, λ, and ρ are free variables. The paper introduces three new distributions as solutions to fractional differential equations, and these are the only invented entities. The axioms are standard tools from fractional calculus, special functions, and probability theory. The heaviest external load is the cited identities from [25] that connect infinite beta products to stable subordinator functionals.

assumptions (5)
  • standard math Fractional integral inversion formula I^α(D^α f)=f for f=I^α g, as per [22, Lemma 2.5]
    Used in uniqueness proofs (Theorems 2.1-2.5) to convert fractional differential equations into fixed point equations.
  • standard math Properties of Barnes double Gamma function and generalized Pochhammer symbols as in [8,23]
    Used throughout Section 4 for Mellin transforms and asymptotic analysis.
  • standard math Moment determinacy via Carleman's criterion and Carlson's theorem
    Used to identify laws from their moments in Section 3 and Proposition 4.3.
  • standard math Existence and uniqueness of the α-stable subordinator with Laplace transform E[e^{-λσ_t}] = e^{-tλ^α}
    Standard Lévy process theory, used for the stochastic representations in Theorems 1.1-1.3.
  • domain assumption The boundary cases α=0 and α=1 are handled with elementary algebraic or differential solutions
    The paper restricts to distribution functions and uses known boundary cases stated in Section 2.
invented entities (3)
  • W_{α,λ,ρ} (fractional Weibull distribution) independent evidence
    purpose: Unique distribution solving D^α_{0+} F = λ x^{ρ-α} \bar F
    Characterized by the differential equation, a stochastic representation, and a series representation via Kilbas-Saigo functions; can be checked numerically or by asymptotics.
  • F_{α,λ,ρ} (fractional Fréchet distribution) independent evidence
    purpose: Unique distribution solving D^α_- \bar F = λ x^{-ρ-α} F
    Characterized by the differential equation, a stochastic representation, and a series representation via Kilbas-Saigo functions; can be checked numerically or by asymptotics.
  • G_{α,λ} (fractional Gumbel distribution) independent evidence
    purpose: Unique distribution solving D^α_+ F = λ^α e^{λx} \bar F
    Characterized by the differential equation and a representation in terms of the Le Roy function; can be checked numerically or by asymptotics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fractional extreme distributions." pith.science (2026). https://pith.science/paper/XOFVOH3U

@misc{pith2026190800584,
  author       = {Pith},
  title        = {Pith review of: Fractional extreme distributions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XOFVOH3U}},
  note         = {Machine review of arXiv:1908.00584}
}
abstract

We consider three classes of linear differential equations on distribution functions, with a fractional order $\alpha\in [0,1].$ The integer case $\alpha =1$ corresponds to the three classical extreme families. In general, we show that there is a unique distribution function solving these equations, whose underlying random variable is expressed in terms of an exponential random variable and an integral transform of an independent $\alpha-$stable subordinator. From the analytical viewpoint, this law is in one-to-one correspondence with a Kilbas-Saigo function for the Weibull and Fr\'echet cases, and with a Le Roy function for the Gumbel case. By the stochastic representation, we can derive several analytical properties for the latter special functions, extending known features of the classical Mittag-Leffler function, and dealing with monotonicity, complete monotonicity, infinite divisibility, asymptotic behaviour at infinity, uniform hyperbolic bounds.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 34 canonical work pages

  1. [25]

    Letemplier and T

    J. Letemplier and T. Simon. On the law of homogeneous sta ble functionals. ESAIM P & S 23, 82-111, 2019

    work page 2019

  2. [1]

    G. E. Andrews, R. Askey and R. Roy. Special functions. Cambridge University Press, Cambridge, 1999

    1999

  3. [2]

    B. C. Arnold, C. A. Robertson and H.-C. Yeh. Some properti es of a Pareto-type distribution. Sankhy¯ a A48 (3), 404-408, 1986

    work page 1986

  4. [3]

    J. H. Barrett. Differential equations of non-integer ord er. Can. J. Math. 6, 529-541, 1954

    work page 1954

  5. [4]

    Berg and J

    C. Berg and J. L. L´ opez. Asymptotic behaviour of the Urba nik semigroup. J. Approx. Theory 195, 109-121, 2015

    work page 2015

  6. [5]

    Bernyk, R

    V. Bernyk, R. C. Dalang and G. Peskir. The law of the suprem um of a stable L´ evy process with no negative jumps. Ann. Probab. 36 (5), 71-101, 2008

    work page 2008

  7. [6]

    J. Bertoin. L´ evy processes.Cambridge University Press, Cambridge, 1996

    work page 1996

  8. [7]

    Bertoin and M

    J. Bertoin and M. Yor. On subordinators, self-similar Ma rkov processes and some factorizations of the exponential variable. Elect. Comm. in Probab. 6, 95-106, 2001

    work page 2001

Show all 36 references
  1. [8]

    Billingham and A

    J. Billingham and A. C. King. Uniform asymptotic expansi ons for the Barnes double gamma function. Proc. Roy. Soc. London Ser. A 453, 1817-1829, 1997

    1997

  2. [9]

    Bondesson

    L. Bondesson. Generalized Gamma convolutions and related classes of dist ributions and densities. Lect. Notes Stat. 76, Springer-Verlag, New York, 1992

    1992

  3. [10]

    Bosch and T

    P. Bosch and T. Simon. On the self-decomposability of th e Fr´ echet distribution.Indag. Math. 24, 626-636, 2013

    2013

  4. [11]

    Carmona, F

    P. Carmona, F. Petit and M. Yor. On the distribution and a symptotic results for exponential functionals of L´ evy processes. In: Exponential functionals and principal values related to Br ownian motion . Bibliot´ eca de la Revista Matem´ atica Iberoamericana, 73-121, 1997

    1997

  5. [12]

    Cs¨ org¨ o, Z

    M. Cs¨ org¨ o, Z. Shi and M. Yor. Some asymptotic properti es of the local time of the uniform empirical process. Bernoulli 5, 1035-1058, 1999

    1999

  6. [13]

    E. C. de Oliveira, F. Mainardi and J. Vaz. Fractional mod els of anomalous relaxation based on the Kilbas and Saigo function. Meccanica 49 (9), 2049-2060, 2014

    2014

  7. [14]

    Dufresne

    D. Dufresne. G distributions and the beta-gamma algebra. Elec. J. Probab. 15, No. 71, 2163-2199, 2010

    2010

  8. [15]

    Flajolet, X

    P. Flajolet, X. Gourdon and P. Dumas. Mellin transforms and asymptotics: Harmonic sums. Theoret. Comput. Sci. 144, 3-58, 1995

    1995

  9. [16]

    Garappa, F

    R. Garappa, F. Mainardi and S. V. Rogosin. On a generaliz ed three-parameter Wright function of Le Roy type. Frac. Calc. Appl. Anal. 20 (5), 1196-1215, 2017

    2017

  10. [17]

    Gorenflo, A

    R. Gorenflo, A. A. Kilbas, F. Mainardi and S. V. Rogosin. Mittag-Leffler functions, related topics and applications. Springer Verlag, Heidelberg, 2014

    2014

  11. [18]

    Gorenflo and F

    R. Gorenflo and F. Mainardi. Fractional calculus and sta ble probability distributions. Arch. Mech. 50, 377-388, 1998

    1998

  12. [19]

    Hirsch, C

    F. Hirsch, C. Profeta, B. Roynette and M. Yor. Peacocks and associated martingales, with explicit constr uctions. Springer-Verlag, Mailand, 2011

    2011

  13. [20]

    L. F. James, B. Roynette and M. Yor. Generalized gamma co nvolutions, Dirichlet means, Thorin measures, with explicit examples. Probab. Surv. 5, 346-415, 2008

    2008

  14. [21]

    Jedidi, T

    W. Jedidi, T. Simon and M. Wang. Density solutions to a cl ass of integro-differential equations. J. Math. Anal. Appl. 458 (1), 134-152, 2018

    2018

  15. [22]

    A. A. Kilbas, H. M. Srivastava and J. J. Trujillo. Theory and applications of fractional differential equation s. North-Holland, Amsterdam, 2006

    2006

  16. [23]

    Kuznetsov

    A. Kuznetsov. On extrema of stable processes. Ann. Probab. 39 (3), 1027-1060, 2011

    2011

  17. [24]

    E. Le Roy. Valeurs asymptotiques de certaines s´ eries p roc´ edant suivant les puissances enti` eres et positives d’une variable r´ eelle.Darboux Bull. 24 (2), 245-268, 1899

  18. [26]

    F. W. J. Olver. Asymptotics and special functions. Academic Press, New York, 1974. 46 L. BOUDABSA, T. SIMON, AND P. V ALLOIS

    1974

  19. [27]

    R. N. Pillai. On Mittag-Leffler functions and related dis tributions. Ann. Inst. Statist. Math. 42 (1), 157-161, 1990

    1990

  20. [28]

    K. Sato. L´ evy processes and infinitely divisible distributions. Cambridge University Press, Cambridge, 1999

    1999

  21. [29]

    Shaked and J

    M. Shaked and J. G. Shanthikumar. Stochastic orders and their applications. Springer Verlag, New York, 2007

    2007

  22. [30]

    T. Simon. Comparing Fr´ echet and positive stable laws. Elec. J. Probab. 19, No. 16, 1-25, 2014

    2014

  23. [31]

    T. Simon. Mittag-Leffler functions and complete monoton icity. Int. Transf. Spec. Funct. 26 (1), 36-50, 2015

    2015

  24. [32]

    F. W. Steutel. Infinite divisibility in theory and pract ice. Scand. J. Statist. 6, 57-64, 1979

    1979

  25. [33]

    C. S. Tapiero and P. Vallois. Fractional randomness. Phys. A 462, 1161-1177, 2016

    2016

  26. [34]

    C. S. Tapiero and P. Vallois. Implied fractional hazard rates and default risk distributions. Probab. Uncertain. Quant. Risk 14, No. 2, 835-843, 2017

    2017

  27. [35]

    C. S. Tapiero and P. Vallois. Fractional randomness and the Brownian Bridge. Phys. A 503, 835-843, 2018

    2018

  28. [36]

    E. C. Titchmarsh. The theory of functions. Oxford University Press, Oxford, 1939. Institut de Math´ematiques, Ecole Polytechnique F´ed´erale de Lausanne, CH-1015 Lausanne, Switzer- land. Email: lotfi.boudabsa@epfl.ch Laboratoire P aul P ainlev´e, UMR 8524, Universit ´e de Lill...

    1939

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.