First passage times of general sequences of random vectors: A large deviations approach (Q1807272): Difference between revisions

Let \(Y_1,Y_2,\dots \) be a sequence of random variables from \(\mathbb R^d.\) For any subset \( A \subset \mathbb R^d \) let \( T(A) = \inf \{n: Y_n \in A\}\) be the first passage time. Limiting behaviour of the normalized first passage time \(T^{\varepsilon}(A) = \varepsilon \inf \{n: \varepsilon Y_n \in A\}\) is studied for \(\varepsilon \to 0.\) It is proved that if the sequence \( \{\frac {1}{n} Y_n\} \) satisfies the large deviation principle and some other regularity conditions on the sequence \(\{Y_n\}\) and on the set \(A\) are satisfied, the probability that \(T^{\varepsilon}(A) \in C \) decays exponentially in \(\varepsilon \) as \(\varepsilon \to 0\) for any closed or open subset \( C \subset [0, \infty).\) Furthermore, it is proved that for any \(\gamma > 0\) \[ P(|T^{\varepsilon}(A) - \rho |> \gamma \mid T^{\varepsilon}(A) < \infty) \to 0 \quad \text{as } \varepsilon \to 0, \] where \(\rho \) is a specified constant, and an analogous result is established for the normalized first passage place. Some results from convex analysis may be of independent interest.