Pith. sign in

REVIEW 5 minor 56 references

On the Existence of Unbiased Hypothesis Tests: An Algebraic Approach

T0 review · 0 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read For multinomial models, an unbiased test exists exactly when a polynomial separates the null hypothesis from the alternative.

desk verdict A genuinely new algebraic criterion for existence of unbiased tests in multinomial models, with a useful threshold concept and constructive methods; the main proof holds up and the scope limitation is acknowledged. read the letter →

arxiv 2506.08259 v1 pith:PTTYL5LF submitted 2025-06-09 math.ST stat.MEstat.TH

Andrew McCormack This is my paper
classification math.STstat.MEstat.TH MSC 62F0362H1713P1014P10
keywords unbiasedtestpowerpolynomialseparatingunbiasednessthresholdGröbnerbasissumofsquarescontingencytablesemialgebraicset
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

The paper proves that, for count data from a multinomial model, the existence of a non-trivial unbiased test is a purely algebraic question: such a test exists at sample size $n$ exactly when some polynomial of degree at most $n$ keeps the null hypothesis on its non-positive side and the alternative on its non-negative side. Since power functions in this setting are homogeneous degree-$n$ polynomials, the minimum sample size that permits an unbiased test is the minimum degree of a separating polynomial, called the unbiasedness threshold. This turns a statistical existence problem into a computational algebra problem, and the paper shows how Gröbner basis and sum-of-squares techniques give upper bounds and often exact thresholds for contingency tables, independence, log-linear, and mixture models. It also exhibits natural hypotheses—ordered categories, mixture polytopes with interior vertices, and log-linear hypotheses with irrational coefficients—that admit no non-trivial unbiased test for any sample size.

What carries the argument

The load-bearing object is the power polynomial $\beta_\phi(\pi)=E_\pi[\phi(X)]$, which in the multinomial model is a homogeneous degree-$n$ polynomial whose coefficients are the test's rejection probabilities on each count configuration. Subtracting the level $\alpha$ and homogenizing produces a separating polynomial $\tilde{\beta}$ of degree at most $n$; the paper works directly with these polynomials instead of test statistics. For algebraic null hypotheses, the separating polynomials live in the vanishing ideal $I_{\mathbb{R}^{k-1}}(P_0)$, and reduced Gröbner bases with respect to graded monomial orders supply the degree data that determine the unbiasedness threshold. Sums of squares $\tilde{\beta}=\sum h_i^2 f_i^2$ built from defining equations give explicit separating polynomials, and the coefficient polytope $C_{n,\alpha}(P_0)$ encodes the box constraints that a polynomial must satisfy to come from an actual test, which is what decides whether a uniformly most powerful unbiased test exists.

What would settle it

Enumerate all randomized test functions for the $2\times2$ independence hypothesis at sample size $n=3$ and check whether any is non-trivial and unbiased; the theory predicts threshold $4$, so finding such a test would refute Theorem 7 and Theorem 1.

Watch

Extended reading notes

Core claim

The central discovery is Theorem 1: in a full multinomial model with a closed null hypothesis $P_0 \subset \Delta_{k-1}$ and alternative $P_A = \Delta_{k-1}\setminus P_0$, a non-trivial unbiased (NTUB) test exists at sample size $n$ if and only if there is a polynomial $\tilde{\beta}$ of degree at most $n$, non-constant on the simplex, with $P_0 \subseteq \{\tilde{\beta}\le 0\}$ and $P_A \subseteq \{\tilde{\beta}\ge 0\}$; for a strictly unbiased (SUB) test the requirements tighten to $P_0=\{\tilde{\beta}\le0\}$ and $P_A=\{\tilde{\beta}>0\}$. The reason is that every power function is a homogeneous degree-$n$ polynomial in the cell probabilities (Lemma 3), and translating by the test level $\alpha$ turns the unbiasedness inequalities into these sub-level set conditions. Consequently the unbiasedness threshold is exactly the least degree of a separating polynomial. The paper draws several consequences: any null hypothesis with a SUB test must be a basic closed semialgebraic set; algebraic null hypotheses always admit SUB tests through sums of squares; polytope null hypotheses have unbiased tests only when all vertices lie on the simplex boundary; and log-linear hypotheses on the natural parameters are testable exactly when all coefficients are rational.

Load-bearing premise

The reduction depends on the multinomial assumption that every power function is a degree-$n$ polynomial; for continuous or infinite sample spaces, power functions are not polynomials and the algebraic criterion does not apply.

Editorial extensions

If this is right

  • Any null hypothesis that admits a strictly unbiased test must be a basic closed semialgebraic set; hypotheses traced by non-algebraic curves, such as the exponential curve in Example 2, have no such test at any sample size.
  • For algebraic null hypotheses a strictly unbiased test always exists, and the thresholds are bounded by $2\max_i\deg f_i$ for any defining equations; Gröbner basis computations refine these bounds and often make them exact.
  • For the hypothesis that a $p\times q$ contingency table has rank less than $r$, both the NTUB and SUB thresholds equal $2r$; in particular, independence in a $2\times2$ table requires at least $4$ observations.
  • The classical UMPU test for a linear hypothesis on multinomial log-odds is non-trivial exactly when the coefficients of the hypothesis are rational; if any coefficient is irrational, the UMPU test is the trivial constant test.
  • UMPU tests can exist for non-exponential families at some sample sizes and disappear at larger ones, so existence depends on both the level $\alpha$ and the sample size.

Reading between the lines

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

  • Editorial extension: the same polynomial-separation criterion should carry over to any discrete sampling scheme whose power function is a polynomial, such as product-multinomial tables with fixed margins, with the separation happening inside the marginal polytope rather than the full simplex.
  • Editorial extension: identifying sample size with polynomial degree suggests that exact unbiased testing has an information-complexity reading—the threshold quantifies how many counts are needed to certify a semialgebraic separation, so thresholds like $2\binom{k}{2}$ for ties among $k$ categories measure the difficulty of the hypothesis itself.
  • Editorial extension: the coefficient-polytope peeling characterization suggests a concrete algorithm for deciding UMPU existence by vertex enumeration, and it leaves open whether the componentwise-maximum-vertex condition is necessary for all sample sizes, not just $n'=1$.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 5 minor

Summary. This paper develops an algebraic characterization of when non-trivial unbiased (NTUB) and strictly unbiased (SUB) hypothesis tests exist for multinomial models on finite sample spaces. The central result (Theorem 1) states that an NTUB test exists at sample size n if and only if there is a polynomial of degree at most n separating the null and alternative hypothesis sets in the sense of sub-level sets, with a strict version for SUB tests. The authors define the unbiasedness threshold as the minimum degree of such a separating polynomial, prove that it equals twice the degree of a generator for principal vanishing ideals under smoothness conditions, and compute it for several classes of models, including contingency tables with bounded rank, log-linear hypotheses, and mixture models. They also study UMPU tests through the coefficient polytope and show that their existence can depend on the level and the sample size. All proofs are in the supplementary material.

Significance. The paper provides a clean and genuinely useful criterion: existence of unbiased tests in a multinomial model is equivalent to a semialgebraic separation condition, and the unbiasedness threshold is the minimum degree of a separating polynomial. The characterization is proven in both directions, the normalization between arbitrary separating polynomials and genuine power polynomials is explicit, and the paper contributes constructive Gröbner-basis and coefficient-polytope methods. For principal ideals the threshold is exactly 2 deg(f) with a unique UMPU test at that sample size; the bounded-rank contingency-table result and the log-linear rationality criterion are new and falsifiable. The main limitation to finite multinomial sample spaces is acknowledged in Section 3.1 and in the conclusion. The verification here found no load-bearing errors.

minor comments (5)
  1. [Section 2, Example 1] The alternative hypothesis is stated as PA={θ1,θ2}, which overlaps the null P0={θ1}; the intended alternative is presumably PA={θ2,θ3}, and the table indeed lists three distributions. Please correct the statement.
  2. [Supplementary Material, proof of Theorem 1] In the 'only if' direction, the translated polynomial uses β−α and claims sup_{π∈P0} β~(π)=0. This is correct only if α is the actual size of the test; otherwise one should subtract the true size s=sup_{π∈P0} β(π). Please clarify this point.
  3. [Supplementary Material, proof of Theorem 3] The step 'the resulting test is the trivial test. As it is the UMPU test, there does not exist an NTUB test for this hypothesis' is terse; it should explicitly invoke Lemma 1, because an NTUB test of an arbitrary size would imply an NTUB test of the nominal level α, contradicting the triviality of the UMPU test.
  4. [Main text and Supplement] The 'Similarity on the boundary' lemma is numbered Lemma 4 in the main text but Lemma 3 in the supplement; please harmonize the numbering.
  5. [Section 1] There are several typographical errors (for example, 'the the' in the first sentence, 'it is utilized' for 'it utilizes', and 'can be can be tuned' in the introduction). A careful proofreading pass is recommended.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the main algebraic characterization is derived from first principles and external tools, not from self-citation or fitted inputs.

full rationale

The derivation is self-contained. Theorem 1 is proved directly from the definition of a power function in Equation (3), the injective linear correspondence between test functions and coefficient-constrained homogeneous polynomials in Lemma 3, and an explicit normalization argument that converts a separating polynomial into a genuine power polynomial. None of these steps presupposes the existence of an unbiased test or the separating polynomial it is meant to establish. The unbiasedness threshold is defined as the minimum degree of a separating polynomial after the equivalence is proved, so the equality is a stated consequence rather than a hidden input. The algebraic results in Sections 5 and 6 use standard outside machinery, such as Groebner bases and the Nullstellensatz, and the paper's own Lemma 5 derives the vanishing ideal from a parameterization rather than assuming it. The UMPU discussion invokes the classical Lehmann-Romano theorem [35] as external evidence, not as a self-citation, and the examples are concrete computations rather than fitted predictions. No load-bearing step reduces by construction to its own assumptions, and no self-citation determines the outcome.

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

The central claims rest on standard algebraic geometry and on modeling assumptions that the null set is closed and, for threshold computations, algebraic. No free parameters are fit to data, and no new entities are postulated.

assumptions (5)
  • domain assumption The null hypothesis set P0 is a closed subset of the probability simplex.
    Needed in Theorem 1 for the existence of SUB tests and for the supremum of the size to be attained. The paper states this explicitly.
  • domain assumption The statistical model is the full multinomial model, so the power function of any test is a homogeneous polynomial of degree n in π.
    This polynomial structure is the foundation of the entire framework, introduced in Section 3.
  • domain assumption For threshold computations, the null hypothesis is algebraic, that is, the zero set of a collection of polynomials.
    Assumed in Section 5 onward, enabling the use of ideals and Gröbner bases.
  • standard math Standard algebraic geometry results, including the Nullstellensatz and Gröbner basis theory, are taken as given.
    Used throughout Sections 5 and 6 as black-box tools.
  • domain assumption In Theorems 5 and 8, regularity conditions hold: f has non-zero gradient on P0, and the Jacobian of the generators has rank m on P0.
    These are technical conditions that make the directional-derivative arguments work; they are stated explicitly in the theorems.

how reviews work

0 comments
Cite this review

Pith. "Pith review of On the Existence of Unbiased Hypothesis Tests: An Algebraic Approach." pith.science (2026). https://pith.science/paper/PTTYL5LF

@misc{pith2026250608259,
  author       = {Pith},
  title        = {Pith review of: On the Existence of Unbiased Hypothesis Tests: An Algebraic Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PTTYL5LF}},
  note         = {Machine review of arXiv:2506.08259}
}
read the original abstract

In hypothesis testing problems the property of strict unbiasedness describes whether a test is able to discriminate, in the sense of a difference in power, between any distribution in the null hypothesis space and any distribution in the alternative hypothesis space. In this work we examine conditions under which unbiased tests exist for discrete statistical models. It is shown that the existence of an unbiased test can be reduced to an algebraic criterion; an unbiased test exists if and only if there exists a polynomial that separates the null and alternative hypothesis sets. This places a strong, semialgebraic restriction on the classes of null hypotheses that have unbiased tests. The minimum degree of a separating polynomial coincides with the minimum sample size that is needed for an unbiased test to exist, termed the unbiasedness threshold. It is demonstrated that Gr\"obner basis techniques can be used to provide upper bounds for, and in many cases exactly find, the unbiasedness threshold. Existence questions for uniformly most powerful unbiased tests are also addressed, where it is shown that whether such a test exists can depend subtly on the specified level of the test and the sample size. Numerous examples, concerning tests in contingency tables, linear, log-linear, and mixture models are provided. All of the machinery developed in this work is constructive in the sense that when a test with a certain property is shown to exist it is possible to explicitly construct this test.

Figures

Figures reproduced from arXiv: 2506.08259 by the authors.

Figure 1
Figure 1. The model P0 Ă ∆2 (black curve) from Example 2. Consider modifying P0 by embedding this set within a facet of ∆3. Specifically, let P0 “ tπ P ∆3 : π “ pθ, expp´θq, 1 ´ θ ´ expp´θq, 0qu be contained in the facet of the simplex with π4 “ 0. The polynomial β˜pπq “ π1 satisfies the condition (4) for an NTUB test. However, for any candidate polynomial β˜ the same argument provided in the previous paragraph shows that β˜p… view at source ↗
Figure 2
Figure 2. Contours of the power function βpπq of the test statistic maxpx1, x2q plotted as a function of pπ1, π2q P r0, 1 2 s 2 for the sample sizes n “ 15 (left) and n “ 40 (right) at increments of 0.05. square. The same argument generalizes for other null hypotheses described by affine constraints that have vertices, or lower-dimensional faces, that are contained in the relative interior of the simplex. See the null hypothe… view at source ↗
Figure 3
Figure 3. The polytope null hypothesis on the left has a SUB test, while the hypothesis on the [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

56 extracted references · 55 canonical work pages

  1. [1]

    C. Adcock. Sample size determination: a review. Journal of the Royal Statistical Society: Series D (The Statistician) , 46(2):261–283, 1997

    work page 1997

  2. [2]

    A. Agresti. Exact inference for categorical data: recent advances and continuing controver- sies. Statistics in medicine , 20(17-18):2709–2722, 2001

    work page 2001

  3. [3]

    Amiri and R

    S. Amiri and R. Modarres. Comparison of tests of contingency tables. Journal of biophar- maceutical statistics, 27(5):784–796, 2017

    work page 2017

  4. [4]

    S. Aoki, H. Hara, and A. Takemura. Markov bases in algebraic statistics . Springer Series in Statistics. Springer, New York, 2012

    work page 2012

  5. [5]

    N. Ay, J. Jost, H. V. Lˆ e, and L. Schwachh¨ ofer.Information geometry, volume 64 of Ergeb- nisse der Mathematik und ihrer Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results in Mathematics and Related Areas. 3rd Series. A Series of Modern Surveys in Mathematics] . Springer, Cham, 2017

    work page 2017

  6. [6]

    S. K. Bar-Lev and B. Reiser. An exponential subfamily which admits UMPU tests based on a single test statistic. Ann. Statist., 10(3):979–989, 1982

    work page 1982

  7. [7]

    V. Barnett. The ordering of multivariate data. J. Roy. Statist. Soc. Ser. A , 139(3):318–355, 1976

    work page 1976

  8. [8]

    V. P. Bhapkar. On tests of marginal symmetry and quasi-symmetry in two and three- dimensional contingency tables. Biometrics, pages 417–426, 1979

    work page 1979

Show all 56 references
  1. [9]

    M. W. Birch. The detection of partial association. II. The general case. J. Roy. Statist. Soc. Ser. B , 27:111–124, 1965

    1965

  2. [10]

    A. H. Bowker. A test for symmetry in contingency tables. Journal of the american statistical association, 43(244):572–574, 1948

    1948

  3. [11]

    L. D. Brown. Fundamentals of statistical exponential families with applications in statistical decision theory, volume 9 of Institute of Mathematical Statistics Lecture Notes—Monograph Series. Institute of Mathematical Statistics, Hayward, CA, 1986

    1986

  4. [12]

    Cheng, F

    Y. Cheng, F. Su, and D. A. Berry. Choosing sample size for a clinical trial using decision analysis. Biometrika, 90(4):923–936, 2003. 24

    2003

  5. [13]

    Christensen

    R. Christensen. Log-linear models and logistic regression . Springer Texts in Statistics. Springer-Verlag, New York, second edition, 1997

    1997

  6. [14]

    J. E. Cohen and U. G. Rothblum. Nonnegative ranks, decompositions, and factorizations of nonnegative matrices. Linear Algebra Appl., 190:149–168, 1993

    1993

  7. [15]

    J. B. Conway. Functions of one complex variable , volume 11 of Graduate Texts in Mathe- matics. Springer-Verlag, New York-Berlin, second edition, 1978

    1978

  8. [16]

    D. Cox, J. Little, and D. OShea. Ideals, varieties, and algorithms: an introduction to computational algebraic geometry and commutative algebra . Springer Science & Business Media, 2013

    2013

  9. [17]

    Derksen and V

    H. Derksen and V. Makam. Maximum likelihood estimation for matrix normal models via quiver representations. SIAM J. Appl. Algebra Geom. , 5(2):338–365, 2021

    2021

  10. [18]

    Diaconis and B

    P. Diaconis and B. Sturmfels. Algebraic algorithms for sampling from conditional distribu- tions. Ann. Statist., 26(1):363–397, 1998

    1998

  11. [19]

    A. Dobra. Markov bases for decomposable graphical models. Bernoulli, 9(6):1093–1108, 2003

    2003

  12. [20]

    M. Drton. Likelihood ratio tests and singularities. Ann. Statist., 37(2):979–1012, 2009

    2009

  13. [21]

    Drton, S

    M. Drton, S. Kuriki, and P. Hoff. Existence and uniqueness of the Kronecker covariance MLE. Ann. Statist., 49(5):2721–2754, 2021

    2021

  14. [22]

    Drton, B

    M. Drton, B. Sturmfels, and S. Sullivant. Lectures on algebraic statistics , volume 39 of Oberwolfach Seminars. Birkh¨ auser Verlag, Basel, 2009

    2009

  15. [23]

    Drton and H

    M. Drton and H. Xiao. Wald tests of singular hypotheses. Bernoulli, 22(1):38–59, 2016

    2016

  16. [24]

    Geiger, C

    D. Geiger, C. Meek, and B. Sturmfels. On the toric algebra of graphical models. Ann. Statist., 34(3):1463–1492, 2006

    2006

  17. [25]

    Gibilisco, E

    P. Gibilisco, E. Riccomagno, M. P. Rogantin, and H. P. Wynn, editors. Algebraic and geometric methods in statistics . Cambridge University Press, Cambridge, 2010

    2010

  18. [26]

    F. A. Graybill and R. D. Morrison. Sample size for a specified width confidence interval on the variance of a normal distribution. Biometrics, 16:636–641, 1960

    1960

  19. [27]

    D. R. Grayson and M. E. Stillman. Macaulay2, a software system for research in algebraic geometry. Available at http://www2.macaulay2.com

  20. [28]

    Gross, S

    E. Gross, S. Petrovi´ c, and D. Stasi. Goodness of fit for log-linear network models: dynamic Markov bases using hypergraphs. Ann. Inst. Statist. Math. , 69(3):673–704, 2017

    2017

  21. [29]

    Gross and S

    E. Gross and S. Sullivant. The maximum likelihood threshold of a graph. Bernoulli, 24(1):386–407, 2018

    2018

  22. [30]

    H. Hara, T. Sei, and A. Takemura. Hierarchical subspace models for contingency tables. J. Multivariate Anal., 103:19–34, 2012

    2012

  23. [31]

    K. F. Hirji. Exact analysis of discrete data . Chapman & Hall/CRC, Boca Raton, FL, 2006. 25

    2006

  24. [32]

    P. Hoff. Smaller p-values via indirect information. J. Amer. Statist. Assoc. , 117(539):1254– 1269, 2022

    2022

  25. [33]

    Hosten and S

    S. Hosten and S. Sullivant. A finiteness theorem for Markov bases of hierarchical models. J. Combin. Theory Ser. A , 114(2):311–321, 2007

    2007

  26. [34]

    S. Kreiner. Analysis of multidimensional contingency tables by exact conditional tests: techniques and strategies. Scand. J. Statist. , 14(2):97–112, 1987

    1987

  27. [35]

    E. L. Lehmann and J. P. Romano. Testing statistical hypotheses. Springer Texts in Statistics. Springer, Cham, fourth edition, 2021

    2021

  28. [36]

    E. L. Lehmann and H. Scheff´ e. Completeness, similar regions, and unbiased estimation. I. Sankhy¯ a, 10:305–340, 1950

    1950

  29. [37]

    Miller and B

    E. Miller and B. Sturmfels. Combinatorial commutative algebra , volume 227 of Graduate Texts in Mathematics . Springer-Verlag, New York, 2005

    2005

  30. [38]

    T. S. Motzkin. The arithmetic-geometric inequality. In Inequalities (Proc. Sympos. Wright- Patterson Air Force Base, Ohio, 1965), pages 205–224. Academic Press, New York-London, 1967

    1965

  31. [39]

    Paindaveine

    D. Paindaveine. On UMPS hypothesis testing. Ann. Inst. Statist. Math. , 76(2):289–312, 2024

    2024

  32. [40]

    V. Powers. Certificates of positivity for real polynomials—theory, practice, and applications, volume 69 of Developments in Mathematics . Springer, Cham, [2021] ©2021

    2021

  33. [41]

    SenGupta

    A. SenGupta. Optimal tests in multivariate exponential distributions. In The exponential distribution, pages 351–376. Gordon and Breach, Amsterdam, 1995

    1995

  34. [42]

    G. Shieh. Sample size calculations for logistic and Poisson regression models. Biometrika, 88(4):1193–1199, 2001

    2001

  35. [43]

    D. F. Signorini. Sample size for poisson regression. Biometrika, 78(2):446–450, 1991

    1991

  36. [44]

    P. W. Smith, J. J. Forster, and J. W. McDonald. Monte carlo exact tests for square contingency tables. Journal of the Royal Statistical Society Series A: Statistics in Society , 159(2):309–321, 1996

    1996

  37. [45]

    Somekh-Baruch, A

    A. Somekh-Baruch, A. Leshem, and V. Saligrama. On the non-existence of unbiased esti- mators in constrained estimation problems. IEEE Trans. Inform. Theory, 64(8):5549–5554, 2018

    2018

  38. [46]

    Sturma, M

    N. Sturma, M. Drton, and D. Leung. Testing many constraints in possibly irregular models using incomplete U -statistics. J. R. Stat. Soc. Ser. B. Stat. Methodol. , 86(4):987–1012, 2024

    2024

  39. [47]

    Sullivant

    S. Sullivant. Algebraic statistics, volume 194 of Graduate Studies in Mathematics. American Mathematical Society, Providence, RI, 2018

    2018

  40. [48]

    C. Uhler. Geometry of maximum likelihood estimation in Gaussian graphical models. Ann. Statist., 40(1):238–261, 2012. 26

    2012

  41. [49]

    Watanabe

    S. Watanabe. Algebraic geometry and statistical learning theory , volume 25 of Cambridge Monographs on Applied and Computational Mathematics . Cambridge University Press, Cambridge, 2009

    2009

  42. [50]

    G. M. Ziegler. Lectures on polytopes , volume 152 of Graduate Texts in Mathematics . Springer-Verlag, New York, 1995. 27 Supplementary Material: Proofs Lemma 1. The four different sets of level- α, unbiased, non-trivial unbiased, and strictly unbi- ased tests are all convex se...

    1995

  43. [51]

    , φk´1q where each φi is a rational function and there exists a point xP W where the denominator of every φi does not vanish at x [16, Def 5.5.4]

    A rational map φ : W Ñ V between two complex varieties W and V is a map φ “ pφ1, . . . , φk´1q where each φi is a rational function and there exists a point xP W where the denominator of every φi does not vanish at x [16, Def 5.5.4]. An irreducible variety V “ VCk´1pfp1q, . . ...

  44. [52]

    Furthermore, each term detprUπVsI,Jq in the symmetrized polynomial is equal to˘ detpπI1,J1q for some I1, J1. Without loss of generality it can be assumed, by grouping all of the terms ˘ detpπI,Jq of the symmetrized polynomial together, that each hI,Jpπq has the property that h...

  45. [53]

    This provides the needed contradiction that hI,J “ 0 for every I, J. The parameterization φpa, bq“p a⊺, 1´1⊺ p´1aq⊺pb⊺, 1´1⊺ q´1bqP Rpˆq can be used in Lemma 5 to find the ideal of the set of set of matrices that have rank one and are contained in affine hull of the probabilit...

  46. [54]

    , I0l0u the set ProjpI01,...,I0l0q ` Cn,αpP0q ˘ has a componentwise maximum elementph˚ I01,

    For Vp0q“t I01, . . . , I0l0u the set ProjpI01,...,I0l0q ` Cn,αpP0q ˘ has a componentwise maximum elementph˚ I01, . . . , h˚ I0l0 q. The vertex set Vp0q is described in Definition 7

  47. [55]

    , Ijlju the set ProjpIj1,...,Ijljq ` Cn,αpP0qXtp hIq : hIj1a“ h˚ Ij1a ,@j1ă j, aď lj1u ˘ set has a componentwise maximum ph˚ Ij1,

    Inductively, for every j with Vpjq“t Ij1, . . . , Ijlju the set ProjpIj1,...,Ijljq ` Cn,αpP0qXtp hIq : hIj1a“ h˚ Ij1a ,@j1ă j, aď lj1u ˘ set has a componentwise maximum ph˚ Ij1, . . . , h˚ Ijlj q. If a UMPU test exists it has the power polynomial f 2h˚` α, where h˚ is the poly...

  48. [56]

    Thus, if a UMPU test exists it must have an h of the form hpπq“ 0.6pπ2 1` π2 2` π2 3q` 1.2pπ1π2` π1π3` π2π3q

    The first, third, and sixth coordinates are uniquely maximized at the last vertex. Thus, if a UMPU test exists it must have an h of the form hpπq“ 0.6pπ2 1` π2 2` π2 3q` 1.2pπ1π2` π1π3` π2π3q. 43

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.