An extension of a logarithmic form of Cramér's ruin theorem to some FARIMA and related processes (Q981000): Difference between revisions

The classical Cramér estimate in its logarithmic form asserts that under appropriate conditions, the maximum \(M\) of a negatively drifted random walk satisfies \(\lim_{t\to+\infty} t^{-1}\log \mathbb P[M>t]=-\theta\) for some explicit constant \(\theta\). The paper under review proves a similar estimate for \((g,F)\)-processes, a class of processes generalizing random walks and fractional ARIMA processes. Given a distribution function \(F\) and a real-analytic function \(g(t)=\sum_{i=0}^{\infty}g_it^i\) defined on \((-1,1)\), a \((g,F)\)-process is defined by \(S_n=\sum_{i=0}^{n-1} g_i X_{n-i}\), where \(\{X_i, i\in \mathbb N\}\) are i.i.d.\ random variables with distribution function \(F\). Under appropriate conditions on \(g\) and \(F\) the authors prove a logarithmic form of the Cramér estimate for the maximum \(M=\max_{n\geq 0} S_n\). Also, the most likely paths as well as the most likely behavior of the innovations leading to an unusually high maximum \(M\) are exhibited.