Semiparametric approach to point source modelling in epidemiology and criminology

By treating the conditional approach that was suggested by Diggle and Rowlingson as a generalized additive model, we provide a semiparametric method for point process modelling with point source interventions. We illustrate the flexibility of this approach with two applications. The first is a reanalysis of an epidemiological case-control data set in which we compare the semiparametric fit with a previously reported parametric model. The second is an application to a complex intervention in the Brazilian city of Belo Horizonte, in which we show how the installation of 60 closed-circuit television cameras has changed the spatial distribution of crimes within an area of high criminal activity.