General Purpose Convolution Algorithm in S4-Classes by means of FFT

Abstract: Object orientation provides a flexible framework for the implementation of the convolution of arbitrary distributions of real-valued random variables. We discuss an algorithm which is based on the discrete Fourier transformation (DFT) and its fast computability via the fast Fourier transformation (FFT). It directly applies to lattice-supported distributions. In the case of continuous distributions an additional discretization to a linear lattice is necessary and the resulting lattice-supported distributions are suitably smoothed after convolution. We compare our algorithm to other approaches aiming at a similar generality as to accuracy and speed. In situations where the exact results are known, several checks confirm a high accuracy of the proposed algorithm which is also illustrated at approximations of non-central $\chi^2$-distributions. By means of object orientation this default algorithm can be overloaded by more specific algorithms where possible, in particular where explicit convolution formulae are available. Our focus is on R package distr which includes an implementation of this approach overloading operator "+" for convolution; based on this convolution, we define a whole arithmetics of mathematical operations acting on distribution objects, comprising, among others, operators "+", "-", "*", "/", and "^".