Workshop - Spatial Statistics and Image Analysis in Biology

Jeudi, 26 Mai, 2016 - 15:15 - 15:40
Janine Illian
CREEM Univ. St Andrews, United Kingdom
The owls are not what they seem - marked point processes from a different perspective

Data structures that detail the locations of objects or events in space, i.e. spatial point patterns, are relevant in a wide range of applications – notably in the context of ecology and environmental sciences but are equally relevant elsewhere, e.g. in geophysics or terrorism studies. It is hence not surprising that there have been many advances in spatial point process methodology recently, enabling increasingly complex models to be formulated and fitted. These capture the spatial structures inherent in point patterns with a focus on making them relevant in applications.
In particular, practical relevance increases when the objects are no longer merely regarded as points in space but as objects with properties. This implies that information on attributes of the objects represented by the points is included in a model in addition to their spatial location. Referred to as “marks”, these data on, e.g., the size or the type of the object may provide an improved understanding of spatial structures if included in an analysis. They may not only help explaining the spatial structure formed by the objects, but may themselves exhibit a spatial structure of interest that varies with the spatial structure of the locations. What is more, this dependence between marks and pattern might be of interest in itself. Hence elements in a model representing that dependence are no longer “nuisance parameters”; their structure is relevant and interpretable in the context of the application.
In this talk we discuss a number of applications where flexible marked point process models are particularly relevant – be it because there is a specific interest in interpreting the mark structure or be it because interpreting a data structure as a marked point pattern may account for complex observation processes.





ANR CNRS Rennes 1 Rennes 2 Nantes INSA Rennes INRIA ENSRennes UBO UBS Angers UBL