No label defined (Q4389717)

In this dissertation a new and interesting approach to the approximation of optimal stopping problems for discrete-time random sequences is given. The sequences are imbedded in a sequence of two-dimensional point processes which are assumed to converge to some more simple limiting process, typically a Poisson-point process or a cluster process. Under some further conditions it is shown that the optimal stopping problem can be approximated by the optimal stopping problem of the limiting point process. In the case of Poisson processes this limiting problem is reduced to some first-order differential equation for the boundary curves. Several interesting applications of this method are given. At first a `general' solution is found for the case of optimal stopping of independent sequences thus extending results of \textit{D. P. Kennedy} and \textit{R. P. Kertz} [Adv. Appl. Probab. 22, No. 2, 396-411 (1990; Zbl 0704.60021), ibid. 24, No. 2, 241-266 (1992; Zbl 0758.60046), Ann. Probab. 19, No. 1, 329-341 (1991; Zbl 0722.60036) and in: Strategies for sequential search and selection in real time. Contemp. Math. 125, 217-230 (1992; Zbl 0751.60042)]. Further applications are given to autoregressive processes, stopping of chain dependent sequences and stopping of maxima.