The Use of the Predictive Posterior of the 3-Parameter Log-Normal Distribution for the Quantification of Measurement Uncertainty in EMC

摘要

In some fields of metrology, such as impulse measurements, the measurement model function, on which the quantification of uncertainty is based, may be highly non-linear. Further, physical constraints determine corresponding constraints on the statistical model, that mandate the use of skewed probability density functions (PDFs). Hence a numerical analysis based on Monte Carlo random sampling is necessary and the standard set of available PDFs (such as the one proposed by the Supplement 1 to the Guide to the expression of Uncertainty in Measurement, the GUMS1) is to be enlarged to include asymmetric and flexible (i.e. adaptable to the problem at hand) PDFs. An application of these concepts to the measurement of the rise-time of an impulse is presented where the use of the 3-parameter lognormal PDF is suggested.