{"id":"8d668616-6df6-449f-aee1-58e3bdc5a822","arxiv_id":"1909.01884","paper_version":3,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The ratio of Laplace transforms of powers of a function is injective on monotone right-analytic functions, and this yields an identification result for auction models with unobserved heterogeneity.","lead":"The paper proves that the ratio of Laplace transforms of two powers of a function determines the function, within a class of smooth, nondecreasing functions. For auction theory, this means the distribution of bidders' idiosyncratic values can in principle be recovered from the distributions of the highest and second-highest bids.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; the central injectivity proof is sound, with only the already-noted display sign error in the auction formula.","rationale":"Both the reader and I find the proof of Theorems 1–2 mathematically sound. The monotonicity/right-analyticity assumptions are exactly the ones used in the bootstrap and are natural in the auction application. I checked the delicate points: Lemma 1's asymptotic comparison of t^{-p}h(t) forces k=ℓ; Lemma 2's coefficient comparison at degree d is valid because ℓ>k so terms with two or more a/b factors have degree >d; Lemma 3's strict inequality I1+I2>0 follows from f>g on (a,a+δ), g(a+δ-u)≥g(u), and the antisymmetry of the combined integrand. The only verified defects are typographical in displays; they do not affect the uniqueness statement. Thus the reader's CONDITIONAL verdict is appropriate but no stronger concern arises.","tokens_in":10506,"tokens_out":25149,"duration_ms":253856,"concrete_test":"Verify the K formula numerically for N=2 and F(x)=1-e^{-θx}. From the order-statistics CDFs, E e^{-λε_(2)} = 2θ²/((λ+θ)(λ+2θ)) and E e^{-λε_(1)} = 2θ/(λ+2θ), so K=θ/(λ+θ). Since H_{1,2}(F,λ)=hat{F²}/hat{F}=2θ/(λ+2θ), the correct relation H/[N-(N-1)H] gives θ/(λ+θ), while the printed H/[N+(N-1)H] gives θ/(λ+3θ). This settles the sign error. Independently, rederive Lemma 1's C1=C2 comparison with b defined as g^{(ℓ)}(0) to confirm the printed 'b:=f^{(ℓ)}(0)' is a typo.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I cannot identify a load-bearing flaw in the central claim. Theorem 2 is proved with exactly the assumptions it needs: Lemmas 1 and 2 transfer all right derivatives at 0; Lemma 3 uses monotonicity for the inequality g(a+δ-u) ≥ g(u) that makes Q(a+δ)>0; right analyticity then extends equality from a neighborhood of 0 to all of [0,∞). The one step in Lemma 2 that looks delicate on a first reading—the assertion that any term containing f1 or g1 is o(t^d)—is justified because f1(x)=o(x^ℓ) gives a uniform bound |f1(ut)| ≤ ε t^ℓ for t small, uniformly in u∈[0,1]. The definite errors I find are confined to displays: the K(F,λ) formula in Theorem 3 has the wrong sign (the denominator should be N-(N-1)H_{N-1,N}(F,λ), not N+(N-1)H_{N-1,N}(F,λ)), and in Lemma 1 the constant b should be g^{(ℓ)}(0), not f^{(ℓ)}(0). Neither error enters the proof of injectivity.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the injectivity of the map f ↦ H_{n,m}(f,·) = \\widehat{f^n}/\\widehat{f^m} on suitable classes of functions on [0,∞). It proves that for distinct positive integers m,n the ratio determines f uniquely among polynomials (up to global sign when n-m is even) and among nonnegative nondecreasing càdlàg functions that are right analytic at every point and of exponential order. The proof compares the small-time asymptotics of the convolution f^n * g^m, uses beta-function computations to transfer all right derivatives at 0, and then bootstraps equality from a right neighborhood of 0 to the whole line via a comparison lemma. The paper also applies the result to an auction model, where F is the common CDF of idiosyncratic shocks and H_{N-1,N}(F,·) is shown to determine F, so that the ratio of Laplace transforms of the two highest bids identifies F.","tokens_in":10692,"tokens_out":24319,"duration_ms":216028,"significance":"The main uniqueness result (Theorem 2) is a genuine contribution: it provides a positive answer to a natural inverse problem for nonlinear Laplace-transform ratios under explicit and checkable hypotheses. The proof is self-contained and, modulo the display errors listed below, rigorous; it uses only beta-function asymptotics and real analytic continuation, with no parameter fitting. The paper is also honest in stating the open conjecture without monotonicity, and the polynomial/entire-function cases are cleanly separated. If the displayed errors are corrected, the auction application gives a useful identification result for a model with unobserved heterogeneity.","major_comments":[{"comment":"The displayed simplification of the beta-function ratio is incorrect because the factor (km)!/(kn)! is omitted. The ratio equals n/m times (km)!/(kn)! times (ℓ-k+kn)!/(ℓ-k+km)!; after restoring this factor, the conclusion that the ratio is >1 for n>m and <1 for n<m still follows (because ℓ>k), so the argument can be repaired, but the printed equality is false.","section":"Section 2, Lemma 2 (equation after (8))"},{"comment":"The formula K(F,λ) = H_{N-1,N}(F,λ)/(N+(N-1)H_{N-1,N}(F,λ)) is wrong. Combining (14) and (15) gives K = \\widehat{F^N}/(N\\widehat{F^{N-1}}-(N-1)\\widehat{F^N}) = 1/(N H_{N-1,N}(F,λ)-(N-1)). The conclusion of Theorem 3 survives because the corrected map H ↦ 1/(N H - (N-1)) is injective, but the displayed equality must be fixed.","section":"Section 3, proof of Theorem 3 (formula for K)"},{"comment":"The lognormal CDF is not right analytic at 0; all right derivatives of F at 0 vanish, so the Taylor series of F at 0 is identically zero, which cannot represent F on any right neighborhood of 0. Consequently Theorem 3, as stated with right analyticity at every point of [0,∞), does not apply to the lognormal family, and the claim 'By Theorem 3, the mapping (μ,σ) ↦ K(μ,σ,·) is injective' is unsupported. The authors should either replace the example with a distribution satisfying the theorem's assumptions (e.g., a Weibull distribution with shape parameter 2) or clarify and weaken the regularity condition at 0 and prove the needed variant.","section":"Section 3, Example 2"}],"minor_comments":[{"comment":"In the definition of the constants, b := f^{(ℓ)}(0) should read b := g^{(ℓ)}(0), as b is used in (5) for the ℓ-th derivative of g.","section":"Section 2, Lemma 1"},{"comment":"The comparison of p1 and p2 contains a garbled identity; the intended contradiction is that if p1>p2 then t^{-p2}h(t) = t^{p1-p2} t^{-p1}h(t) tends to 0, contradicting the nonzero limit C2.","section":"Section 2, Lemma 1 proof"},{"comment":"In the sentence after defining i0, the equality f^{(j)}(a+) = f^{(j)}(a+) should read f^{(j)}(a+) = g^{(j)}(a+).","section":"Section 2, Lemma 3 proof"},{"comment":"The denominator in the displayed definition of K(μ,σ,λ) and in (13) is missing the factor λ; it should be E e^{-λε_(N-1)}, not E e^{-ε_(N-1)}.","section":"Section 3, Example 2 and display (13)"}],"recommendation":"minor_revision","confidential_remarks":"The lognormal example seems to have escaped the reader's report; it is a substantive error in the application section, although it does not affect the core uniqueness theorem. Please ensure the authors fix it, for instance by choosing a distribution that is right analytic at 0. The paper is otherwise acceptable after local corrections."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here is my take on arXiv:1909.01884. The headline: the paper proves a real injectivity result for the ratio of Laplace transforms of powers of a function, and the auction identification corollary is legitimate, but the printed text has a few typos that should be fixed before anyone cites it.\n\nWhat is actually new: the map f -> \\hat{f^n}/\\hat{f^m} is injective on monotone right-analytic functions, on polynomials (up to sign when n-m is even), and on entire functions. The proof strategy is solid: beta-function asymptotics at zero transfer all derivatives at 0, a monotonicity argument gives a sign contradiction at any point where f and g first disagree, and right analyticity then pushes equality to the whole half-line. Lemmas 1 through 3 and Theorems 1 and 2 are coherent and convincing. The auction application is a clean reduction: knowing the ratio of Laplace transforms of the two highest order statistics of the idiosyncratic errors is equivalent to knowing H_{N-1,N}(F,·), so Theorem 2 yields identification of the CDF. That answers a question that had been posed in the econometrics literature.\n\nThe soft spots are all in the presentation, not the substance. In Lemma 1 the proof defines b := f^{(ℓ)}(0) where it should be g^{(ℓ)}(0). In Lemma 2's ratio computation, the simplification drops the factor (km)!/(kn)!; the displayed product equality is numerically false, though the conclusion that the ratio is >1 when n>m still holds for the correct expression. In Theorem 3, the formula for K(F,λ) is misprinted: with H_{N-1,N}(F,λ) = \\hat{F^{N-1}}/\\hat{F^N}, the correct expression is K = 1/(N H_{N-1,N} - (N-1)), not H_{N-1,N}/(N + (N-1)H_{N-1,N}). The identification conclusion survives because K is an injective function of H, so this typo is not load-bearing, but as printed it is wrong.\n\nThe authors honestly state the conjecture that monotonicity can be removed, which is the main boundary of the result. No fitted parameters, no circular dependencies. The paper deserves a serious referee; the typos are easily corrected and the central theorem holds up. I would accept it for peer review and ask for the display corrections before publication.","headline":"A genuine injectivity result for ratios of Laplace transforms of powers, with an auction-theory identification corollary; the core proofs are sound, but several display typos need correction.","tokens_in":11238,"tokens_out":6818,"would_cite":true,"duration_ms":54832,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["44A10","26Axx","91B70","91B26"],"pacs":[],"model":"deepseek-v4-flash","headline":"Laplace transform ratios of powers identify a nondecreasing function","keywords":["Laplace transform","ratio uniqueness","order statistics","auction theory","right analytic functions","monotone functions","identification problem","inverse Laplace transform"],"falsifier":"Construct two smooth, nonnegative, nondecreasing, positive-on-(0,∞) functions f and g, right analytic on [0,∞), with L($f^{2}$)L(g) = L($g^{2}$)L(f) but f ≠ g; such a pair would disprove Theorem 2. Since no such pair is known, the theorem's correctness hinges on the absence of this counterexample.","tokens_in":10278,"feed_emoji":"📈","tokens_out":4573,"duration_ms":44681,"temperature":0.7,"pith_summary":"This paper asks whether the ratio of Laplace transforms of two distinct powers of a function f determines f uniquely. It proves that for nonnegative, nondecreasing, cadlag, right-analytic functions of exponential order, the ratio H_{n,m}(f, λ) = L(f^n)(λ) / L(f^m)(λ) completely identifies f. For polynomials and entire functions, the same conclusion holds, with the only ambiguity being a global sign when n − m is even. The result matters because this exact ratio appears in auction theory, where it encodes the distribution of bidder-specific private values from the two highest bids.","feed_headline":"Laplace transform ratios of powers identify a nondecreasing function","feed_subtitle":"A ratio of two Laplace transforms of powers of f determines f exactly, settling an auction-theory identification problem.","key_machinery":"The central object is the ratio H_{n,m}(f, λ) = L(f^n)(λ) / L(f^m)(λ), an analytic quantity for large real λ when f is of exponential order. The proof's machinery is a bootstrap: Lemmas 1 and 2 compare Taylor coefficients at 0 via $\\beta$-function integrals of convolutions f^n * g^m and f^m * g^n, showing all right derivatives at 0 match. Lemma 3 then takes a point a where f = g up to a, and uses monotonicity to force a sign contradiction if the first differing right derivative had opposite sign, so right analyticity lets equality step from a to a neighborhood beyond a. This pushes equality from 0 to all of [0,∞).","core_discovery":"The central claim is Theorem 2: if f and g are nonnegative, nondecreasing, cadlag, right analytic at every point of [0,∞), of exponential order, and strictly positive for x > 0, then equality of H_{n,m}(f, ·) and H_{n,m}(g, ·) forces f = g. The proof first shows that all right derivatives of f and g at 0 coincide, using Taylor expansions and $\\beta$-function integrals applied to the convolution identity f^n * g^m = f^m * g^n. It then considers the largest a such that f and g agree on [0, a); Lemma 3 shows that if they agree up to a, then their right derivatives at a also agree, so right analyticity extends the agreement past a, contradicting maximality unless a = ∞.","pith_inferences":["If the authors' conjecture that monotonicity can be dropped is correct, the auction identification result would hold for all distribution functions, not only nondecreasing smooth ones.","The same ratio-injectivity phenomenon may hold for other integral transforms with a convolution structure, where ratios of powers act as enough information to determine the underlying function up to symmetry.","The bootstrap suggests a numerical recovery scheme for f from H_{n,m}, though the reliance on derivatives at 0 likely makes the reconstruction sensitive to small errors in the ratio.","When n − m is odd, an explicit inversion formula analogous to the inverse Laplace transform may exist, since all Taylor coefficients are then determined without sign ambiguity."],"forward_implications":["Any two functions in the stated class with the same ratio of Laplace transforms of powers are the same function, so the map f ↦ H_{n,m}(f, ·) is injective on that class.","In the symmetric auction model with unobserved common value and idiosyncratic shocks, the common distribution F of the idiosyncratic component is identified from the Laplace transforms of the highest and second-highest bids (Theorem 3).","For polynomials and entire functions, equality of the ratio identifies f exactly when n − m is odd; when n − m is even, the only possible ambiguity is f versus −f.","The derivative-matching argument provides a constructive route: the Taylor data of f at 0 are recovered from H_{n,m}, and the bootstrap then determines f on the whole half-line."],"supporting_citations":[{"why":"Supplies the definition and extension facts for right-analytic functions used throughout the proof.","marker":"[6]"},{"why":"Provides the inverse Laplace transform used for the base case where one of the exponents is zero and for identifying f^n.","marker":"[14]"},{"why":"Presents the auction/order-statistics model that motivates the question and inspires Lemma 3's comparison argument.","marker":"[10]"},{"why":"Justifies analyticity of the Laplace transforms of powers for sufficiently large real parts of λ.","marker":"[3]"}],"fun_headline_variants":["Ratios of Laplace transforms recover the function uniquely","Power-transform ratio pins down the function","Identification from Laplace transform ratios of powers","Laplace ratio of powers determines the function","How a ratio of Laplace transforms identifies a function"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof needs the functions to be nondecreasing; without monotonicity, the sign comparison in Lemma 3 collapses, so equality cannot be pushed from a neighborhood of zero to the whole half-line.","fun_headline_variants_meta":{"raw":{"variants":["Ratios of Laplace transforms recover the function uniquely","Power-transform ratio pins down the function","Identification from Laplace transform ratios of powers","Laplace ratio of powers determines the function","How a ratio of Laplace transforms identifies a function"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000529,"raw_usage":{"total_tokens":2498,"prompt_tokens":843,"completion_tokens":1655,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":459,"completion_tokens_details":{"reasoning_tokens":1589}},"tokens_in":459,"tokens_out":1655,"duration_ms":11141,"temperature":1.0,"reasoning_tokens":1589,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T05:07:54.425540+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct two smooth, nonnegative, nondecreasing, positive-on-(0,∞) functions f and g, right analytic on [0,∞), with L($f^{2}$)L(g) = L($g^{2}$)L(f) but f ≠ g; such a pair would disprove Theorem 2. Since no such pair is known, the theorem's correctness hinges on the absence of this counterexample.","supporting_citations":[{"cited_title":"and Parks, H","cited_arxiv_id":null,"evidence_quote":"Supplies the definition and extension facts for right-analytic functions used throughout the proof."},{"cited_title":"The Laplace Transform","cited_arxiv_id":null,"evidence_quote":"Provides the inverse Laplace transform used for the base case where one of the exponents is zero and for identifying f^n."},{"cited_title":"and Xiao, R","cited_arxiv_id":null,"evidence_quote":"Presents the auction/order-statistics model that motivates the question and inspires Lemma 3's comparison argument."},{"cited_title":"Introduction to the Theory and Application of the Laplace Tr ansformation","cited_arxiv_id":null,"evidence_quote":"Justifies analyticity of the Laplace transforms of powers for sufficiently large real parts of λ."}],"review_version":1}