On the relationship between micro and macro correlations in nuclear measurement uncertainties

Consideration is given to the propagation of micro correlations between the component experimental errors (corresponding to diverse attributes of the measurement process) through to the macro correlations between the total errors in the final derived experimental values. Whenever certain micro correlations cannot be precisely specified, the macro correlations must also be uncertain. However, on the basis of fundamental principles from mathematical statistics, it is shown that these uncertainties in the macro correlations can be substantially smaller than the individual uncertainties for specific micro correlations, provided that the number of distinct attributes contributing to the total experimental error is reasonably large. Furthermore, the resulting macro correlations are shown to be approximately normally distributed regardless of teh distributions assumed for the micro correlations. Examples are provided to demonstrate these concepts and to illustrate their relevance to experimental nuclear research. (orig.)