Composite likelihood inference for a time series of spatial point patterns

Abdollah Jalilian (University of Glasgow)

Wednesday 26th November 14:00-15:00
Maths 311B

Abstract

In some applications data come as spatial point patterns observed repeatedly over time, for example the locations of trees recorded every few years in a tropical rainforest. Understanding the spatial and temporal dynamics in such point patterns involves modelling the appearance (birth) and disappearance (death) of points between time periods. In this talk, a composite likelihood approach will be presented for making inference when the full likelihood is too complex to compute. The method relies on conditional models that use only first-order properties of the data, providing a simpler and efficient way to estimate model parameters with minimal assumptions about spatial and temporal dependence structure. I will also discuss how asymptotic results can be derived for short time series covering large spatial regions and how robust covariance estimators can be constructed, using rainforest census data as an example.

 

Related publication:

Jalilian, A., Cuevas-Pacheco, F., Xu, G., & Waagepetersen, R. (2025). Composite likelihood inference for space-time point processes. Biometrics, 81(1), ujaf009. https://doi.org/10.1093/biomtc/ujaf009

Add to your calendar

Download event information as iCalendar file (only this event)