2022

Kuronen, Mikko

Spatial point patterns arise in a number of applications from different disciplines. They represent locations of objects or events of interest. Such data is analysed and modelled using point process statistics. This work develops new statistical models and methods for challenges encountered in a few specific applications in forestry and medicine. We consider methods for the analysis of datasets that include artefacts or missing data, introduce new point process models, and suggest tests having graphical interpretation. In one of the applications, we develop models for sweat gland activation data, which is important in early screening of diabetes. To this end, we suggest methods to handle erroneously detected points in the data produced by image analysis. We also consider modelling how the locations of tree seedlings are affected by large trees. Here we propose a Bayesian inference method for handling nonlinear covariates in a log Gaussian Cox process. Furthermore, we present an estimator for forest characteristics in data obtained by terrestrial laser scanning. The new estimator accounts for unobserved trees behind other trees. Finally, we suggest a test with a graphical interpretation for including particular covariates in a point process model.