Differences between statistical unertainty and modeling uncertainty are briefly discussed. It is pointed out that, when different models are proposed for the interpretation of reality, the uncertainty cannot be described in terms of mean value and coefficient of variation. The important question is: which of the proposed models is more reliable than the others?
The attention, then, is concentrated on the comparison between different models proposed for the estimate of the required quantity, looking for a criterion leading to the selection of the most reliable one.
A criterion for comparison of different models is suggested. In the example of application considered in this paper, it proved to be effective, so that the continuation of numerical experiments, exploring different and more complex situations, seems promising.