pith. sign in

arxiv: 2604.25108 · v2 · submitted 2026-04-28 · 🧮 math.PR · math.CO

Terminal Defects, Growing Multiplicity, and Variance Extremality in the Double Dixie Cup Problem

Christopher D. Long This is my paper

Pith reviewed 2026-05-07 15:44 UTC · model grok-4.3

classification 🧮 math.PR math.CO
keywords coupon collector problemdouble dixie cupvariance extremalitypoissonizationgamma reverse hazardmonotone likelihood ratioGumbel limitterminal defect
0
0 comments X

The pith

The variance of time to collect m full coupon sets is minimized exactly when all probabilities are equal, and increases along any deviation from uniformity.

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

The paper establishes that for any number of coupon types N at least 2 and any target multiplicity m at least 1, the waiting time T_m(N) to obtain m complete sets has variance that is uniquely smallest under the uniform probability vector and strictly larger for any other positive probabilities. The proof works exactly for finite N by Poissonizing the process so that the waiting time becomes the maximum of independent Erlang random variables, then comparing the radial derivative of the distribution function to a size-biased version via a monotone-likelihood-ratio argument that relies on a log-scale monotonicity property of the Gamma reverse hazard rate. This settles the finite-variance extremality conjecture of Doumas and Papanicolaou and simultaneously yields a growing-multiplicity Gumbel limit law together with explicit inverse-gamma-tail asymptotics when probabilities are equal. A reader cares because many practical sampling and coverage problems are sensitive to the variability of completion times, and the result identifies the probability vector that minimizes that variability without requiring large-N approximations.

Core claim

For every m ≥ 1 and N ≥ 2, among all positive coupon probability vectors p, the variance of T_m(N) is uniquely minimized at the uniform vector; moreover the variance is strictly increasing along every ray from the uniform vector. The same terminal-defect framework produces a growing-multiplicity Gumbel theorem with expectation and variance asymptotics on the inverse gamma-tail scale in the equal-probability case, recovering the fixed-m variance asymptotic previously conjectured for m=1 and extending it to m = m_N, while also giving endpoint-Laplace limits for power-law probabilities.

What carries the argument

Terminal-defect method after Poissonization, representing completion time as the maximum of independent Erlang variables and comparing the radial derivative of its distribution to the size-biased law via monotone-likelihood-ratio argument based on log-scale monotonicity of the Gamma reverse hazard rate.

Load-bearing premise

The log-scale monotonicity property of the Gamma reverse hazard rate holds and permits the monotone-likelihood-ratio comparison between the radial derivative and the size-biased law after Poissonization.

What would settle it

A concrete probability vector p for small fixed m and N where direct computation of Var(T_m(N)) yields a value smaller than the uniform case, or where the radial derivative fails the monotone-likelihood-ratio ordering against the size-biased law.

read the original abstract

We develop a terminal-defect method for the double Dixie cup problem and use it to prove the finite-variance extremality conjecture of Doumas and Papanicolaou. For every \(m\ge1\) and \(N\ge2\), among all positive coupon probability vectors \(p=(p_1,\ldots,p_N)\), the variance of the time \(T_m(N)\) to collect \(m\) complete sets is uniquely minimized at the uniform vector. We prove the stronger radial statement that the variance is strictly increasing along every ray from the uniform vector. The proof is finite-\(N\) and exact: after Poissonization, the completion time is a maximum of independent Erlang variables, and the radial derivative of its distribution is compared to a size-biased law using a monotone-likelihood-ratio argument based on a log-scale monotonicity property of the Gamma reverse hazard. The same framework gives a growing-multiplicity Gumbel theorem in the equal-probability case, with expectation and variance asymptotics on the inverse gamma-tail scale. This recovers the fixed-\(m\) equal-probability variance asymptotic stated as Conjecture 1 by Doumas and Papanicolaou, classically known for \(m=1\), and extends the mechanism to \(m=m_N\). We also illustrate the unequal-probability theory with endpoint-Laplace limits for power-law probabilities.

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

0 major / 3 minor

Summary. The paper proves the finite-variance extremality conjecture of Doumas and Papanicolaou for the double Dixie cup problem: for every m ≥ 1 and N ≥ 2, among all positive probability vectors p, the variance of the waiting time T_m(N) to collect m complete sets of N coupons is uniquely minimized at the uniform vector, and moreover is strictly increasing along every ray emanating from the uniform vector. The argument is exact and finite-N: after Poissonization the completion time is the maximum of independent Erlang random variables; the radial derivative of its law is compared to a size-biased version via a monotone-likelihood-ratio ordering that follows from an explicit log-scale monotonicity property of the Gamma reverse hazard rate, verified by direct differentiation. The same framework yields a growing-multiplicity Gumbel limit theorem in the uniform case (recovering the m=1 variance asymptotic and extending it to m = m_N) together with endpoint-Laplace limits for power-law probabilities.

Significance. If the central claims hold, the work supplies the first exact finite-N resolution of the variance-extremality conjecture, strengthens it to a radial monotonicity statement, and extends the classical equal-probability asymptotics to growing multiplicity without post-hoc fitting. The proof is self-contained, relying only on elementary properties of Poisson processes, Erlang maxima, and an explicit sign analysis of the Gamma reverse-hazard derivative; this explicit verification is a clear strength that avoids circularity or parameter tuning.

minor comments (3)
  1. §2.3, after Eq. (8): the phrase 'terminal defect' is introduced without a forward reference to its precise definition in §3.1; a one-sentence pointer would improve readability for readers unfamiliar with the coupon-collector literature.
  2. Figure 1 caption: the vertical axis label 'radial derivative of Var' is not accompanied by the scaling factor used in the plot; adding the precise normalization (e.g., multiplied by N^2) would make the visual comparison with the theoretical sign claim immediate.
  3. Theorem 4.2: the statement of the Gumbel limit is given for m_N → ∞ with m_N = o(log N), but the proof sketch in §4.3 only treats the case m_N / log N → 0; a brief remark on the boundary regime would clarify the full range of applicability.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of our manuscript and the recommendation to accept. The report accurately captures the main contributions: the exact finite-N proof of the variance-extremality conjecture via Poissonization and Gamma reverse-hazard analysis, the stronger radial monotonicity property, and the extensions to growing-multiplicity Gumbel limits and power-law endpoint behavior.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The derivation proceeds from Poissonization of the coupon collector process, representing T_m(N) as the maximum of independent Erlang random variables, followed by an explicit verification (via direct differentiation and sign analysis of the resulting expression) of the log-scale monotonicity of the Gamma reverse hazard rate. This monotonicity is then used to establish an MLR ordering between the radial derivative of the distribution and the size-biased law, which signs the radial derivative of the variance. All steps are self-contained within the manuscript, rely only on elementary properties of Poisson processes, Gamma densities, and CDFs, and contain no fitted parameters, self-referential definitions, load-bearing self-citations, or imported uniqueness theorems. The central claim is therefore independent of its inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard properties of Poisson point processes, Erlang distributions as maxima, and the log-scale monotonicity of the Gamma reverse hazard rate; no free parameters or new entities are introduced.

axioms (2)
  • standard math After Poissonization the completion time is the maximum of independent Erlang random variables.
    Invoked to obtain an exact representation of T_m(N) for finite N.
  • standard math The Gamma reverse hazard rate is monotone on the log scale, enabling an MLR comparison with the size-biased law.
    Used to sign the radial derivative of the variance.

pith-pipeline@v0.9.0 · 5539 in / 1510 out tokens · 50070 ms · 2026-05-07T15:44:42.557715+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

9 extracted references · 9 canonical work pages

  1. [1]

    R. K. Brayton, On the asymptotic behavior of the number of trials necessary to complete a set with random selection,Journal of Mathematical Analysis and Applications7(1963), 31–61

    work page 1963

  2. [2]

    A. V. Doumas and V. G. Papanicolaou, The coupon collector’s problem revisited: Generalizing the double Dixie cup problem of Newman and Shepp,ESAIM: Probability and Statistics20 (2016), 367–399

    work page 2016

  3. [3]

    Erd˝ os and A

    P. Erd˝ os and A. R´ enyi, On a classical problem of probability theory,Magyar Tud. Akad. Mat. Kutat´ o Int. K¨ ozl.6(1961), 215–220

    work page 1961

  4. [4]

    Flatto, Limit theorems for some random variables associated with urn models,Annals of Probability10(1982), 927–934

    L. Flatto, Limit theorems for some random variables associated with urn models,Annals of Probability10(1982), 927–934

    work page 1982

  5. [5]

    Kaplan, A generalization of a result of Erd˝ os and R´ enyi,Journal of Applied Probability14 (1977), 212–216

    N. Kaplan, A generalization of a result of Erd˝ os and R´ enyi,Journal of Applied Probability14 (1977), 212–216. 26

    work page 1977

  6. [6]

    D. J. Newman and L. Shepp, The double Dixie cup problem,American Mathematical Monthly 67(1960), 58–61

    work page 1960

  7. [7]

    N. M. Temme, The asymptotic expansion of the incomplete gamma functions,SIAM Journal on Mathematical Analysis10(1979), no. 4, 757–766

    work page 1979

  8. [8]

    N. M. Temme, Asymptotic inversion of incomplete gamma functions,Mathematics of Com- putation58(1992), no. 198, 755–764

    work page 1992

  9. [9]

    Yu, Minimum variance in the coupon collector’s problem,Journal of Applied Probability 54(2017), 655–656

    Y. Yu, Minimum variance in the coupon collector’s problem,Journal of Applied Probability 54(2017), 655–656. 27

    work page 2017