Abstract
This paper presents a simple, neo-classical, method of uncertainty analysis based on the idea that measurement is a procedure that incurs errors. Following a correct interpretation of CIPM Recommendation INC-1 (1980), the method approaches the task of combining random and systematic (deterministic) errors by considering the systematic errors to have been drawn from hypothetical populations whose variances contribute to the law of propagation of error in the usual manner. In Type B evaluation of uncertainty, (which typically relates to systematic components), the method propagates the shapes of the error distributions via the concept of kurtosis and avoids the artificial use of degrees of freedom — which is a known weakness of the procedure of the Guide to the Expression of Uncertainty in Measurement. The method has the properties of internal consistency and transferability, and its coherent theoretical basis promotes universality. The method is tuned for simplicity and good performance in simulated measurements. Such simulations give meaning to the idea that a method of uncertainty analysis is to exhibit validity.