A novel analytical scheme to compute the \(n\)-fold convolution of exponential-sum distribution functions (Q1888285)

The aim of this paper is to develop a novel scheme for the general analytical problem of computing the \(n\)-fold convolution of exponential-sum distribution functions. In a previous paper by \textit{N.-Y. Ma} and \textit{R. P. King} [ibid. 133, No.~1, 83--91 (2002; Zbl 1026.60003)], this problem was solved by applying a complete multinomial expansion theorem and Laplace transform techniques, with the price of rather complex derivations. The analytical solution proposed in this paper reduces the complexity of computation, is simpler to understand conceptually, and easier to be implemented computationally. The first step of the new approach is to reduce the considered problem of computing the \(n\)-fold convolution of exponential-sum distribution functions to the computation of some simple recursive integrals. Then, these recursive integrals are reconstructed with series of delta functions to avoid separations of integrations. Next, the task of solving the obtained recursive integrals can be achieved depending on specific objectives of the problem. If the goal is to compute the \(n\)-fold convolution, it may be sufficient to use the solutions of the recursive integrals by direct integrations, with the help of Maple-like symbolic integration packages. When it is desirable to obtain the general solution patterns of the resulted recursive integrals, the authors develop a general pattern of Laplace transform of the recursive integrals, followed by performing an inverse Laplace transform operation, necessary to obtain the computing formulas for the \(n\)-fold convolution of exponential-sum distribution functions.