Pointwise density estimation on metric spaces and applications in seismology

G. Cleanthous,Athanasios G. Georgiadis,P. A. White
DOI: https://doi.org/10.1007/s00184-024-00948-2
IF: 0.96
2024-02-14
Metrika
Abstract:We are studying the problem of estimating density in a wide range of metric spaces, including the Euclidean space, the sphere, the ball, and various Riemannian manifolds. Our framework involves a metric space with a doubling measure and a self-adjoint operator, whose heat kernel exhibits Gaussian behaviour. We begin by reviewing the construction of kernel density estimators and the related background information. As a novel result, we present a pointwise kernel density estimation for probability density functions that belong to general Hölder spaces. The study is accompanied by an application in Seismology. Precisely, we analyze a globally-indexed dataset of earthquake occurrence and compare the out-of-sample performance of several approximated kernel density estimators indexed on the sphere.
statistics & probability
What problem does this paper attempt to address?