Probabilistic arithmetic. I: Numerical methods for calculating convolutions and dependency bounds (Q916336)

This paper investigates a numerical method for calculating the appropriate convolution of probability distributions, which arises in probabilistic arithmetic calculation. First, several definitions of concepts used in this paper, such as quasi-inverse, copulas and Fréchet bounds, and triangle functions \(\tau\), \(\rho\) and \(\sigma\)-convolutions, are given. Then, the dependency bounds for the four arithmetic operations of addition, subtraction, multiplication, and divisions are derived and their properties are explained with illustrating examples. Algorithms for calculating the dependency bounds and \(\sigma\)-convolutions numerically are developed based on the numerical representation of probability distributions. Some extensions to the above results are also presented. The general idea developed in this paper is closely related to interval arithmetic and fuzzy arithmetic.