Uncertainty propagation by sampling with NumPy: beyond the linear rule

Afterwards you can propagate uncertainties by sampling with NumPy, compare the result with the linear rule, and report a skewed one as median and 95 % interval.