I will be giving the Statistics Seminar at UCLA on October 18th. The talk is titled: “Perfect simulation of repulsive point processes”.
Abstract Entities in competition for resources, such as towns in a country or trees in a forest, often are further apart from each other than would be predicted if their spatial distribution was independently uniform. Repulsive point processes are used to model such data, but rarely are these distributions given in closed form. Instead, they can be given as either an algorithmic model, or as a unnormalized density with respect to a spatial Poisson point process. In this talk I will show how to generate random samples exactly from an algorithmic setup called the Matérn Type III model, and from an unnormalized density called the Strauss model. Also, I will discuss effective ways to turn these samples into estimates using these models.