Some problems of local hitting, scaling, and conditioning (Q996734): Difference between revisions

The present paper gives an informal and non-technical introduction to some of the results in three cases: the simple point processes, the local time random measures and the Dawson-Watanabe superprocesses. A simple point process can be identified with the associated random counting measure \(\xi\), where \(\xi(B)\) gives the number of points in the set \(B\). The expression ``\(\xi\) hits \(B\)'' means that \(\xi(B)>0\). There exist some natural questions: 1. What is the probability that \(\xi\) hits \(B\)? 2. What is the conditional distribution of \(\xi\) outside (or inside) \(B\) given that \(\xi\) hits \(B\)? 3. What is the conditional distribution of \(\xi\) inside \(B\) given that \(\xi\) to be known outside \(B\)? Much of the classical theory of point processes is concerned with finding answers to these questions, or rather some asymptotic results as \(B\) approaches a single point \(s\in S\), often under simplifying assumptions such as stationarity or the existence of moments. The results for point processes are usually not applicable to more general random measures, and entirely new methods can be required. Nevertheless, many of the questions for point processes still make sense: 1. Are the hitting probabilities \(P\{\xi(b)>0\}\) still related to the intensity measure \(E\xi\)? 2. Can we still approximate the Palm distributions by elementary conditional distributions \(P\{\xi\in\cdot| \xi(B)>0\}\)? 3. Is there a duality relationship that connects the interior and exterior conditioning? Here only the last question has a general answer, in the form of a duality theory. As for remaining questions, nice answers, similar to those in the point process case, are available in two cases of special interest, namely, for (A) Local time random measures associated with (suitably regular) regenerative processes; (B) Dawson-Watanabe superprocesses, the most basic class of measure-valued diffusion processes. It is still not clear whether the results for those two classes of processes are special cases of some general theory that remains to be uncovered. The paper is divided into the following 6 sections: I. Introduction. II. Palm measures and duality. III. Simple point processes. IV. Regeneration and local time. V. Measure-valued branching processes. VI. Conclusion and open problems. See also the following monograph of the author: [\textit{O. Kallenberg}, Probabilistic symmetries and invariance principles. New York, NY: Springer (2005; Zbl 1084.60003)].