Back to Volume
Paper: Validation of Statistical Models for Spatial Point Patterns
Volume: 371, Statistical Challenges in Modern Astronomy IV
Page: 22
Authors: Baddeley, A.
Abstract: Statistical methods for analysing a spatial point pattern are reviewed and criticised, with emphasis on the statistical ‘validity’ of the analysis. Three main approaches are to extract a numerical summary of the pattern (such as the pair correlation function), to search for interesting features (such as clusters), or to fit a statistical model to the pattern. We give a statistical critique of each approach, and describe recent improvements, such as inhomogeneous K-functions, LISA statistics, Cox models and model-fitting algorithms. The key question is how to check the ‘validity’ of an analysis. We describe the new concept of residuals for a spatial point process, and its uses in model validation.
Back to Volume