Currently the calculator seems to always assume a normal distribution. For many usecases, a linear one would be more useful or even needed. A normal distribution can even be approximated by multiplying many linear distributions, but I don't know any way to do the reverse.