On computational formulas for densities and moments of compound distributions (Q1586596)

Let \((X_k)\), \(k\geq 1\), be a sequence of i.i.d.\ r.v. Let \(N\) be a counting variate independent of \((X_k)\). Let \(Y=X_1+X_2+\cdots +X_N\). \textit{G. E. Willmot} and \textit{H. H. Panjer} [Insur. Math. Econ. 6, 43-56 (1987; Zbl 0613.62020)] and \textit{B. Sundt} [ASTIN Bull. 22, 61-80 (1992)] found the computational formulae for the density of \(Y\) when the distribution of \(N\) satisfies the first-order linear difference equation \[ (an+1)p_n=[b(n-1)+c]p_{n-1}, \quad n\geq 1, \] or the first-order quadratic difference equation \[ [n(n-1)+an]p_n=[b(n-1)(n-2)+c(n-1)+d]p_{n-1}, \quad n\geq 1. \] The authors give the recurrence formulae for the density of \(Y\) when the distribution of \(N\) belongs to the Ord family, i.e. such that \(p_n={\mathbf P}(N=n)\) satisfies \[ [\alpha n^2+\beta n+\gamma]p_n=[\alpha n^2+(\beta-1) n+\gamma+\delta]p_{n-1}, \quad n\geq 1. \] Moreover, the frequency distributions satisfying \[ (n+c)p_n=(an+ac+b)p_{n-1}, \quad n\geq n_0, \] are also considered. Many examples for concrete distributions are given.