Hahn polynomials for hypergeometric distribution

Plamen Iliev, Yuan XuAdv. Appl. Math 2022 · DOI

TL;DR

This paper introduces and studies a new family of orthogonal polynomials, called Hahn polynomials for the hypergeometric distribution, defined on discrete polyhedral domains. These polynomials generalize classical Hahn polynomials and have applications in probability theory and quantum mechanics, offering a deeper understanding of their structure and properties, including factorization and generating functions.

What Problem Does This Paper Address?

The paper addresses the challenge of defining and understanding orthogonal polynomials for the multivariate hypergeometric distribution, which is a probability distribution on discrete polyhedral domains. While classical orthogonal polynomials like Hahn polynomials are well-studied, their extension to multivariate settings and specific distributions like the hypergeometric distribution requires new theoretical frameworks. The authors aim to bridge this gap by constructing these new polynomials, exploring their structural properties, and connecting them to existing theories. Specifically, they investigate how negative integer parameters in classical Hahn polynomials lead to these new families and how their properties, such as vanishing on specific lattice points, relate to the geometry of the polyhedral domains.

Methods

The core of the paper involves extending the theory of classical Hahn polynomials to a multivariate setting relevant to the hypergeometric distribution. Key techniques include:

  • Analytic Continuation: The authors leverage the fact that many identities for classical Hahn polynomials involve rational functions of their parameters. This allows them to extend these identities to cases where parameters become negative integers, a crucial step for defining the new polynomials.
  • Gram-Schmidt Process: This standard orthogonalization technique is used to construct the orthogonal polynomials with respect to the defined inner product.
  • Factorization Techniques: The paper explores the factorization of these polynomials, which is essential for understanding their vanishing properties on specific lattice points and for relating them to simpler polynomial structures.
  • Generating Functions and Recurrence Relations: The authors derive generating functions and recurrence relations for these new polynomials, which are standard tools for characterizing orthogonal polynomial families and provide insights into their structure.
  • Combinatorial Proofs: The paper relies on detailed combinatorial arguments to prove key results, particularly concerning the structure of index sets and the properties of the polynomials.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Generalization of Hahn Polynomials: The paper successfully constructs a new family of orthogonal polynomials that generalize classical Hahn polynomials to the context of the multivariate hypergeometric distribution on polyhedral domains.
  • Role of Negative Integer Parameters: A significant insight is how setting classical Hahn polynomial parameters to negative integers directly leads to orthogonal polynomials for the hypergeometric distribution. This provides a concrete mechanism for their construction.
  • Connection to Polyhedral Geometry: The structure of these new polynomials is intimately linked to the geometry of the discrete polyhedral domains on which they are defined. Their vanishing properties are directly related to the lattice points within these domains.
  • Factorization Reveals Structure: The factorization of these polynomials is shown to be a powerful tool for understanding their behavior, particularly how they vanish on specific sets of lattice points. This reveals non-trivial structures that are not immediately apparent from their definition.
  • Multivariate Complexity: The paper highlights the increased complexity of these polynomials in higher dimensions compared to the one-variable case, especially when dealing with factorization and the relationship between index sets and domain structures.

Key Results

  • Definition of Hahn Polynomials for Hypergeometric Distribution: The paper formally defines these new polynomials and establishes their orthogonality with respect to the hypergeometric distribution on polyhedral domains.
  • Factorization Properties: Several theorems and propositions detail the factorization of these polynomials, particularly in the two-variable case, revealing linear factors and irreducible components.
  • Generating Functions and Recurrence Relations: Explicit formulas for generating functions and recurrence relations are derived, providing essential tools for their analysis.
  • Examples of Vanishing Polynomials: The paper provides concrete examples of non-trivial polynomials that vanish on large subsets of lattice points within the polyhedral domains, illustrating the theoretical findings.

Limitations and Future Work

  • Complexity in Higher Dimensions: While the paper extends the theory to multiple variables, the complexity of factorization and structural analysis increases significantly with dimension. The paper notes that analogous factorizations to the one-variable case do not always hold in higher dimensions.
  • Irreducibility Conjectures: The paper poses a conjecture regarding the irreducibility of certain polynomials, suggesting that further work is needed to fully characterize their algebraic properties.
  • Further Exploration of Spectral Properties: The paper introduces bispectral properties and difference operators, hinting at deeper connections to quantum integrable systems and suggesting avenues for future research in this area.
  • Positive Definiteness of Linear Functionals: The paper notes that the moment functional associated with some of these polynomials is not positive definite, indicating a departure from standard orthogonal polynomial theory and a potential area for further investigation.

Summary updated

BibTeX
@article{iliev2022hahn,
  author = {Plamen Iliev and Yuan Xu},
  title = {{Hahn polynomials for hypergeometric distribution}},
  journal = {Adv. Appl. Math.},
  year = {2022},
  doi = {10.1016/j.aam.2022.102364},
}
Figure