7. Conclusion

At the end of this article, we can see that there is no change in metrological approach between the analytical and numerical methods. Despite the many criticisms, in many cases the analytical method remains highly relevant, since cases of non-linearity are rare in the practical orders of magnitude of the sources of uncertainty. In complex processes, however, the numerical method is the only viable option. What's more, once the numerical model has been established, it's easy to make a modification, unlike with analytical GUM, where the whole process has to be redone. For a defined process, the Monte Carlo method is therefore more flexible. This method is not all that complicated to implement, and also makes it possible to document the type of probability density law that best represents the set of values of the output variable(s). In many studies, therefore, both methods are feasible, and...