scientific article; zbMATH DE number 176271

From MaRDI portal
Publication:4035378





[For the entire collection see Zbl 0752.00047.] Two different estimation problems for Poisson point processes are discussed. First, a sequence of i.i.d. Poisson processes \(\Phi_ 1,\Phi_ 2,\dots\) on a general measurable space \((X,{\mathcal X})\) is considered which have a finite intensity measure \(\Lambda\). It is shown that the arithmetic mean \(\widehat\Lambda_ n=n^{-1}(\Phi_ 1+\cdots+\Phi_ n)\) is uniformly consistent on some classes of subsets of \(X\) (of Vapnik-Chervonenskis type). The asymptotic minimax efficiency of \(\widehat\Lambda_ n\) in the sense of Le Cam is established for a general class of loss functions. Furthermore, the intensity \(\lambda=\mathbb{E} \Phi([0,1])\) of a homogeneous Poisson point process on the real line (i.e. \(X=\mathbb{R}\)) is estimated on the basis of samples from a certain stochastic process \(\bigl\{X(t)\bigr\}\) with \(X(t)=\int f(t-s) \Phi(ds)\), where the response function \(f\) is unknown. Using a special form of the expectation and covariance function of \(\bigl\{X(t)\bigr\}\), an estimator for \(\lambda\) is constructed which is consistent and asymptotically normal if \(\bigl\{X(t)\bigr\}\) can be sampled on an unboundedly increasing interval.




Cited in
(44)








This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4035378)