ON ESTIMATING THE PERIOD OF A CYCLIC POISSON PROCESS ROELOF HELMERSI WAYAN MANGKU CW I Bogor Agricultural University We propose and investigate a simple nonparametric estimator of the period of a cyclic Poisson process. It is assumed that only a single realization of the Poisson process is observed in a bounded window. We prove consistency and establish a rate of convergence of the proposed estimator when the size of the window expands. 1. Introduction and main result Let X denote a cyclic Poisson point process defined on a probability space ( ,*4, P), with absolutely continuous σ finite mean measure μ w.r.t. Lebesgue measure v and with (unknown) locally integrable intensity func tion λ: R —> R + U {0}, i.e., for any bounded Borel set B, we have μ(B) = J B X(s)ds < oo. In addition, λ is cyclic (with period r), i.e., for some r G R+ (1.1) X(s + kr) = X(s) for all s G R and k E TL. The period r is assumed to be unknown. Suppose that, for some ω G , a single realization X(ω) of the Pois son point process X is observed, though only in a bounded interval (called window) W C R. Since λ is locally integrable, the Poisson point process X always places only a finite number of points in any bounded subset of R. In order to investigate the consistency of an estimator of τ we let the window W depend on "time" n = 1, 2,..., in such a way that \W n \ — • oo, as n — • oo, where \W n \ denotes the size (or Lebesgue measure) of W n . In this set up, a necessary condition for the existence of a consistent estimator (of r) is that J R X(s) ds = EX(R) = oc, which implies that P almost surely the point pattern X{u) contains infinitely many points (cf. Rathbun and Cressie, 1994). Note that for cyclic λ the requirement J^X(s)ds = oo is automatically satisfied, provided the global intensity θ — τ~ ι / Q r X(s) ds of the process X is positive. Therefore we will assume throughout that θ > 0. The aim of this paper is to propose and investigate a simple nonpara metric estimator τ n of the period r of a cyclic Poisson process X, using a single realization X(ω) of X, observed in the window W n . Let θ denote the parameter space, τ £ θ, and let Θ be a bounded open interval in R+ , such that no multiple of τ is contained in Θ. Our estimator τ n of r is obtained as follows: for any δ G θ , define