Geometric sums: bounds for rare events with applications: risk analysis, reliability, queueing (Q4345864)

Let \(X_i\), \(i\geq 1\), be i.i.d. nonnegative random variables, and \(\nu\) an integer-valued random variable which does not depend on \((X_i)_{i\geq 1}\) and has a geometric distribution with parameter \(q\). The random variable \(S_\nu= X_1+X_2+ \cdots +X_\nu\) is called a geometric sum, and can, in many applications, be given the interpretation of the time of the first occurrence of an event. This monograph focuses on the behaviour of \(W_q(x)= P(S_\nu \leq x)\) for small values of \(q\), so that the event is actually a rare event. A starting point for the analysis is Rényi's theorem which states that \(\lim_{q\to 0} W_q(x/q) =1- \exp(-x/m)\) if \(m=E X_1<\infty\). The main limiting result in the book, obtained in Chapter 3, is an extension of Rényi's theorem which allows the distribution function of the \(X_i\)'s to depend on \(q\). Applications of this result in queueing and Markov chain theory are also given in Chapter 3. Various upper and lower bounds for \(W_q\) are derived in Chapter 4, while so-called metric bounds are obtained in Chapter 5. The theoretical results of the Chapters 3, 4 and 5 are used to estimate ruin probabilities in mathematical insurance models in Chapter 6, and are applied to reliability models in Chapter 7. Chapters 1 and 2 are introductory.