Ruin probabilities for a regenerative Poisson gap generated risk process (Q635979)

A risk model is considered, where the claim size distribution depends on the last \(k\) interarrival times. The claim surplus process is \(S(t) = \sum_{i=1}^{N(t)} U_i - c t\), where \(N\) is a Poisson process with rate \(\lambda\), \(U_i\) are independent and independent of \(N\), and \(c > 0\) is the premium rate. If \(k\) consecutive interarrival times are larger than \(r\), then the claim size has distribution \(G\), otherwise it has distribution \(F\). Letting \(\tau(u)\) be the time of ruin, the object of interest is the ruin probability \(\psi(u) = P[\tau(u) < \infty]\). The times when a \(G\)-claim occurs are regeneration times. By a straightforward argument, the mean cycle length is calculated. Using arguments from \textit{S. Asmussen, H. Schmidli} and \textit{V. Schmidt} [Adv. Appl. Probab. 31, No.2, 422--447 (1999; Zbl 0942.60033)], the asymptotics of the ruin probability in the heavy-tailed case is found. The asymptotics in the light-tailed case is based on large deviation techniques, and motivated by \textit{P. W. Glynn} and \textit{W. Whitt} [in: Studies in applied probability. Essays in honour of Lajos Takács. Sheffield: Applied Probability Trust, 131--156 (1994; Zbl 0805.60093)]. First, the Laplace transform \(E[\exp\{\alpha S(\omega) + \beta \omega\}]\) is calculated, where \(\omega\) is a regeneration cycle. From that, the logarithmic asymptotics of the ruin probability can be found. Considering the risk model as a Markov additive process, the exact asymptotics of the ruin probability is found by change of measure techniques. In a last section, the time to ruin is studied. For the heavy-tailed case, the conjecture is formulated that the time of ruin follows asymptotically the same law as in the classical model. In the light-tailed case, the rate at which \(\tau(u)/u\) converges in distribution is found. The paper concludes with remarks and extensions.