5. Defining a confidence interval

The question of confidence intervals is often at the heart of metrological problems. In the GUM, it doesn't really arise, since the assumption of normality is assumed, sometimes incorrectly (see Step "Simulating a measurement model"). In GUM S1, things are different, since we have as many possible measurand values as we have simulations. It's in these simulated values that we'll have to look for the confidence interval of interest, knowing of course that there are many intervals containing the chosen quantity of simulated possible values (often 95%, by convention in metrology) of the measurand.