the goal is to have a function that takes as arguments two discrete distributions and outputs the distribution corresponding to the sum of an independent sample from each (so the pmf is the convolution of the individual ones). My current implementation is using dicts to represent the distributions and is quite inefficient (especially on distributions whose support covers a large range of values). Is there a canonical way to do this with the scipy rv_discrete class? I'd rather not reinvent the wheel.
I want to be exact and not use FFTs, though maybe I should just do that. Thanks for your advice.
