IMPROVING MEASUREMENT UNCERTAINTIES BY USING WEIGHTED MEANS

摘要

The mean value of a group of standards is assumed to be more stable than the value of any single standard, and so this is universally used to give high confidence to a set of measurements. The usual method to obtain the value of an unknown standard is to measure its differences from all the units in a group to predict a new value for the unknown. After repeated measurements over a period of months or years, some standards are found to have values that fit very closely to a linear rate of change with time, whilst others are found to fit less well. It is proposed here that less notice or weight should be given to any standard in a group that varies a lot and more weight should be given to those standards which vary little. Three ways of doing this are proposed and explained.