Rounding in approximations #3
Description
Consider the calculation of getY() in FailureRate. Due to the approximative way we calculate the density, small values in the density function (derived from the reliability function) are not >0 but approximated to 0.0. FailureRate handles this scenario by knowing that as long as reliability != 1, the density cannot be 0.0 for our assumptions and, thus, returns Double.NaN.
I fear this might not be consistently integrated in all classes where this error might occur and the question is where to resolve this issue. I think the better place to resolve this would be the functions that do the approximation, i.e., DensityFunction in this case.
Question to discuss: Do we want to rely on propagating Double.NaN values or should we throw exceptions?