loss-of-significance is a phenomenon occuring in numerical programs when the result is significantly less accurate than the resolution of the floating point numbers used. Often, but not always, loss-of-significance can be circumvented by making the required adjustments to the numerical algorithm.
