The uncertainties module returns its result with the uncertainty specified by linear
error propagation theory, correctly taking into account any direct correlations between variables. This is in fact what Pint uses under the hood. If you need still more advanced approaches to
propagation of uncertainty, Lebigot recommends trying
soerp (second-order approximations) and
mcerp (Monte-Carlo approach).
Lebigot recommends two related uncertainty calculation Python packages if you need still more advanced approaches to
propagation of uncertaintyy but we have not tried them ourselves yet :
soerp (second-order approximations) and
mcerp (Monte-Carlo approach).