Identification of COVID-19 mortality patterns in Brazil by a functional QR decomposition analysis

Abstract

This paper introduces a functional extension of the QR decomposition of linear algebra and shows its application to identify independent patterns (curves) of COVID-19 mortality in Brazil's states. The problem is treated as a subset selection one, and regions of influence of each pattern are then determined by fitting the mortality curves of the remaining states to the main independent ones. Three main patterns are detected: (1) a two-peak curve in central and southern states, (2) a curve with an early single peak concentrated in the Amazonian state of Roraima, and (3) a curve with and early peak and a large recent increase in the Amazonian, northeastern and southeastern states.