Limit theorems for strongly mixing stationary random measures (Q1174263): Difference between revisions

Let \(\zeta_ T\), \(T>0\), be stationary random measures on \([0,1]\) satisfying the strong mixing condition that \(\alpha_{T,\delta_ T}\to 0\) for some values \(\delta_ T\to 0\), where \[ \alpha_{T,\delta}=\sup\{| P(A\cap B)-P(A)P(B)|: A\in {\mathcal B}^ T(0,s), B\in {\mathcal B}^ T(u,1)\} \] the sup also under the restriction that \(0<s<s+\delta\leq u\leq 1\), where \({\mathcal B}^ T(I)\) is the \(\sigma\)-algebra generated by \(\zeta_ T\) over the interval \(I\). Conditions are identified under which \(\zeta_ T @>d>> \zeta\) for a random measure \(\zeta\). Necessarily, the limit is stationary, has no fixed atoms and has independent increments, so that its Laplace functional has the form \[ L_ \zeta(f)=\exp\left[-\alpha\int_ 0^ f(x)dx-\int_ 0^ 1\int_ 0^ \infty(1-e^{- yf(x)})\nu(dy)dx\right] \] for some \(\alpha >0\) and Lévy measure \(\nu\). In particular, \(\zeta_ T @>d>> \zeta\) if and only if there are an interval \(I\) and a random variable \(\eta_ I\) such that \(\zeta_ T(I) @>d>> \eta_ I\), and if and only if the measures \(k_ T\int_ A(1- e^{-x})p_ T(dx)\) converge vaguely to the measure \(\alpha\varepsilon_ o(A)+\int_ A(1-e^{-x})\nu(dx)\), where the \(k_ T\) are appropriately chosen integers converging to \(\infty\). The case that \(\zeta\) is a compound Poisson process is examined in detail. The results are applied to describe the limiting behavior of the exceedance measures associated with a stationary stochastic process \((\xi_ T)\). In this case, \[ \zeta_ T(B)=\int_ 0^ \infty 1(x/T\in B)1(\xi_ x>u_ T)dx \] for the thresholds \(u_ T\to\infty\).