Discrete statistical models with rational maximum likelihood estimator

@article{Duarte2019DiscreteSM,
  title={Discrete statistical models with rational maximum likelihood estimator},
  author={Eliana Duarte and Orlando Marigliano and Bernd Sturmfels},
  journal={arXiv: Statistics Theory},
  year={2019}
}
A discrete statistical model is a subset of a probability simplex. Its maximum likelihood estimator (MLE) is a retraction from that simplex onto the model. We characterize all models for which this retraction is a rational function. This is a contribution via real algebraic geometry which rests on results due to Huh and Kapranov on Horn uniformization. We present an algorithm for constructing models with rational MLE, and we demonstrate it on a range of instances. Our focus lies on models… 

Figures from this paper

Estimating linear covariance models with numerical nonlinear algebra
TLDR
The maximum likelihood degree and its dual analogue are studied, and a new software package LinearCovarianceModels.jl is introduced for solving the score equations.
Logarithmic Voronoi cells
We study Voronoi cells in the statistical setting by considering preimages of the maximum likelihood estimator that tessellate an open probability simplex. In general, logarithmic Voronoi cells are
Families of polytopes with rational linear precision in higher dimensions
In this article, we introduce a new family of lattice polytopes with rational linear precision. For this purpose, we define a new class of discrete statistical models that we call multinomial staged
Maximum Likelihood Estimation from a Tropical and a Bernstein--Sato Perspective
In this article, we investigate Maximum Likelihood Estimation with tools from Tropical Geometry and Bernstein--Sato theory. We investigate the critical points of very affine varieties and study their
Staged tree models with toric structure
Optimal Transport to a Variety
TLDR
This work considers the setting of finite state spaces and describes the solution depending on the choice of the ground metric and the given distribution in terms of the solutions of a finite number of systems of polynomial equations.
Representation of Context-Specific Causal Models with Observational and Interventional Data
TLDR
It is shown that the context-specific information encoded by a CStree can be equivalently expressed via a collection of DAGs, and a global Markov property for CStrees is obtained which leads to a graphical criterion of model equivalence for C Strees generalizing that of Verma and Pearl for DAG models.
D-Modules and Holonomic Functions
TLDR
This lecture notes address the more general case when the coefficients are polynomials, and focuses on left ideals, or D-ideals, which represent holonomic functions in several variables by the linear differential equations they satisfy.
Beyond Linear Algebra
TLDR
The title challenges the reader to venture beyond linear algebra in designing models and in thinking about numerical algorithms for identifying solutions and the role of nonlinear algebra in the study of linear PDE with constant coefficients.
...
...

References

SHOWING 1-10 OF 16 REFERENCES
Equations defining probability tree models
Likelihood Geometry
We study the critical points of monomial functions over an algebraic subset of the probability simplex. The number of critical points on the Zariski closure is a topological invariant of that
Varieties with maximum likelihood degree one
We show that algebraic varieties with maximum likelihood degree one are exactly the images of reduced A-discriminantal varieties under monomial maps with finite fibers. The maximum likelihood
Linear precision for parametric patches
TLDR
The connection between linear precision for toric surface patches and maximum likelihood degree for discrete exponential families in algebraic statistics, and how iterative proportional fitting may be used to compute toric patches is established.
Conditional independence and chain event graphs
Equivalence classes of staged trees
TLDR
A complete characterization of the statistical equivalence classes of CEGs and of staged trees is given and it is shown that all graphical representations of the same model share a common polynomial description.
A Dynamic Programming Algorithm for Learning Chain Event Graphs
TLDR
This paper describes a dynamic programming algorithm for exact learning of chain event graphs from multivariate data that allows reasonably fast approximations and provides clues for implementing more scalable heuristic algorithms.
...
...