AAU Student Projects - visit Aalborg University's student projects portal
A master's thesis from Aalborg University
Book cover


Covariance between a Random Variable and a Point Process

Translated title

Kovarians mellem en tilfældig variabel og en punkt proces

Author

Term

4. term

Publication year

2025

Submitted on

Abstract

Afhandlingen undersøger, hvordan man kvantificerer sammenhængen mellem en spatial punktproces (fx træers positioner) og tilknyttede tilfældige variable, både binære og kontinuerte (fx jordens pH). Ud fra kovariansen mellem punktprocessen og variablen afledes to sumfunktioner, én for binære og én for kontinuerte variable, og der udvikles ikkeparametriske kernel-estimatorer både med og uden kantkorrektion. Estimatorernes ydeevne vurderes i en simuleringsundersøgelse, der omfatter både uafhængige og afhængige scenarier for Poisson- og log-Gaussiske Cox-processer. Simulationerne indikerer, at kantkorrigerede estimatorer klarer sig bedst og kan identificere både tilstedeværelse og fravær af korrelation mellem punktmønstre og variable. Metoderne anvendes endvidere på et datasæt med træer og jordmålinger fra Barro Colorado Island i Panama. Afhandlingen giver også en introduktion til grundlæggende teori om punktprocesser, herunder intensitet, par-korrelationsfunktion, Poisson- og Cox-processer samt kovarians mellem to punktprocesser.

This thesis investigates how to quantify the association between a spatial point process (e.g., tree locations) and attached random variables that are either binary or continuous (e.g., soil pH). From the covariance between the point process and the variable, two summary functions are derived—one for binary and one for continuous variables—and nonparametric kernel estimators are developed with and without edge correction. The estimators’ performance is assessed in a simulation study covering independent and dependent scenarios for Poisson point processes and log-Gaussian Cox processes. Simulations indicate that edge-corrected estimators perform best and can detect both the presence and absence of correlation between point patterns and variables. The methods are further applied to a dataset of trees and soil measurements from Barro Colorado Island in Panama. The thesis also reviews key point process theory, including intensity, pair correlation, Poisson and Cox processes, and covariance between two point processes.

[This abstract was generated with the help of AI]