Geometry of Log-Concave Density Estimation

Shape-constrained density estimation is an important topic in mathematical statistics. We focus on densities on $$\mathbb {R}^d$$Rd that are log-concave, and we study geometric properties of the maximum likelihood estimator (MLE) for weighted samples. Cule, Samworth, and Stewart showed that the logarithm of the optimal log-concave density is piecewise linear and supported on a regular subdivision of the samples. This defines a map from the space of weights to the set of regular subdivisions of the samples, i.e. the face poset of their secondary polytope. We prove that this map is surjective. In fact, every regular subdivision arises in the MLE for some set of weights with positive probability, but coarser subdivisions appear to be more likely to arise than finer ones. To quantify these results, we introduce a continuous version of the secondary polytope, whose dual we name the Samworth body. This article establishes a new link between geometric combinatorics and nonparametric statistics, and it suggests numerous open problems.

[1]  J. D. Loera,et al.  Triangulations: Structures for Algorithms and Applications , 2010 .

[2]  R. Samworth,et al.  Smoothed log-concave maximum likelihood estimation with applications , 2011, 1102.1191.

[3]  M. An.  Log-Concave Probability Distributions: Theory and Statistical Testing , 1996 .

[4]  Richard J. Samworth,et al.  Maximum likelihood estimation of a multi-dimensional log-concave density: Estimation of a Multi-dimensional Log-concave Density , 2010 .

[5]  Robert B. Gramacy,et al.  Maximum likelihood estimation of a multivariate log-concave density , 2010 .

[6]  K. I. Gross,et al.  Total positivity, spherical series, and hypergeometric functions of matrix argu ment , 1989 .

[7]  J. Wellner,et al.  Estimation of a convex function: characterizations and asymptotic theory. , 2001 .

[8]  Jesús A. De Loera,et al.  The polytope of all triangulations of a point configuration , 1996, Documenta Mathematica.

[9]  L. Duembgen,et al.  Maximum likelihood estimation of a log-concave density and its distribution function: Basic properties and uniform consistency , 2007, 0709.0334.

[10]  Karim A. Adiprasito,et al.  A Geometric Lower Bound Theorem , 2015, 1507.06638.

[11]  Alexander I. Barvinok,et al.  Computing the volume, counting integral points, and exponential sums , 1992, SCG '92.

[12]  Rekha R. Thomas.  Lectures in Geometric Combinatorics , 2006, Student mathematical library.

[13]  U. Grenander.  On the theory of mortality measurement , 1956 .

[14]  Martin L. Hazelton,et al.  Assessing log-concavity of multivariate densities , 2011 .

[15]  Mark Yuying An,et al.  Logconcavity versus Logconvexity: A Complete Characterization , 1998 .

[16]  I. M. Gelʹfand,et al.  Discriminants, Resultants, and Multidimensional Determinants , 1994 .

[17]  G. Walther.  Inference and Modeling with Log-concave Distributions , 2009, 1010.0305.

[18]  B. Sturmfels,et al.  Combinatorial Commutative Algebra , 2004 .