Score matching estimators for directional distributions

One of the major problems for maximum likelihood estimation in the well-established directional models is that the normalising constants can be difficult to evaluate. A new general method of "score matching estimation" is presented here on a compact oriented Riemannian manifold. Important applications include von Mises-Fisher, Bingham and joint models on the sphere and related spaces. The estimator is consistent and asymptotically normally distributed under mild regularity conditions. Further, it is easy to compute as a solution of a linear set of equations and requires no knowledge of the normalizing constant. Several examples are given, both analytic and numerical, to demonstrate its good performance.

[1]  Leif Ellingson,et al.  Nonparametric Statistics on Manifolds and Their Applications to Object Data Analysis , 2015 .

[2]  Steffen Lauritzen,et al.  Linear estimating equations for exponential families with application to Gaussian linear concentration models , 2013, 1311.0662.

[3]  Tomonari Sei,et al.  Calculating the normalising constant of the Bingham distribution on the sphere using the holonomic gradient method , 2013, Stat. Comput..

[4]  R Core Team,et al.  R: A language and environment for statistical computing. , 2014 .

[5]  Andrew T. A. Wood,et al.  Saddlepoint approximations for the normalizing constant of Fisher--Bingham distributions on products of spheres and Stiefel manifolds , 2013 .

[6]  K. Mardia.  Statistical approaches to three key challenges in protein structural bioinformatics , 2013 .

[7]  Kanti V. Mardia,et al.  Mixtures of concentrated multivariate sine distributions with applications to bioinformatics , 2012 .

[8]  S. Lauritzen,et al.  Proper local scoring rules , 2011, 1101.5011.

[9]  S. R. Jammalamadaka,et al.  Directional Statistics, I , 2011 .

[10]  K. Mardia,et al.  Maximum likelihood estimation using composite likelihoods for closed exponential families , 2009 .

[11]  Kanti V. Mardia,et al.  A multivariate von mises distribution with applications to bioinformatics , 2008 .

[12]  Jürgen Jost,et al.  Riemannian Geometry and Geometric Analysis, 5th Edition , 2008 .

[13]  Aapo Hyvärinen,et al.  Some extensions of score matching , 2007, Comput. Stat. Data Anal..

[14]  Steffen L. Lauritzen,et al.  The Geometry of Decision Theory , 2006 .

[15]  Aapo Hyvärinen,et al.  Estimation of Non-Normalized Statistical Models by Score Matching , 2005, J. Mach. Learn. Res..

[16]  K. Mardia,et al.  Directions and projective shapes , 2005, math/0508280.

[17]  Harshinder Singh,et al.  Probabilistic model for two dependent circular variables , 2002 .

[18]  S. R. Jammalamadaka,et al.  Topics in Circular Statistics , 2001 .

[19]  S. Rosenberg.  The Laplacian on a Riemannian Manifold: The Laplacian on a Riemannian Manifold , 1997 .

[20]  J. Jost.  Riemannian geometry and geometric analysis , 1995 .

[21]  I. Chavel.  Eigenvalues in Riemannian geometry , 1984 .

[22]  J. Kent.  The Fisher‐Bingham Distribution on the Sphere , 1982 .

[23]  K. Mardia,et al.  A general correlation coefficient for directional data and related regression problems , 1980 .

[24]  Rudolf Beran,et al.  Exponential Models for Directional Data , 1979 .

[25]  E. Stein,et al.  Introduction to Fourier Analysis on Euclidean Spaces. , 1971 .