Nonparametric estimation of highest density regions for COVID-19

TL;DR

This paper compares classical and hybrid nonparametric methods for estimating highest density regions, which help identify COVID-19 hot-spots, through simulations and real data analysis.

Contribution

It introduces a comparative analysis of existing and new hybrid algorithms for estimating high-density regions in the context of COVID-19 data.

Findings

Hybrid method performs better in simulations.

Both methods successfully identify COVID-19 hot-spots.

Application to real data demonstrates practical utility.

Abstract

Highest density regions refer to level sets containing points of relatively high density. Their estimation from a random sample, generated from the underlying density, allows to determine the clusters of the corresponding distribution. This task can be accomplished considering different nonparametric perspectives. From a practical point of view, reconstructing highest density regions can be interpreted as a way of determining hot-spots, a crucial task for understanding COVID-19 space-time evolution. In this work, we compare the behavior of classical plug-in methods and a recently proposed hybrid algorithm for highest density regions estimation through an extensive simulation study. Both methodologies are applied to analyze a real data set about COVID-19 cases in the United States.