A Method of Evaluating Discrepant Data

Abstract：

In present paper several methods of evaluating discrepant data are introduced and compared each other briefly. The advantage and disadvantage of these methods(WM: weighted mean, LRSW: Limitation of relative statistical weights, MBAYS: Modified Bayesian Technique, NR: Normalized Residuals, and RA: Rajeval technique) are outlined. On the basis of these analysis and comparison, a new method i.e. Double-Mean method (DM) of evaluating discrepant data was proposed. The Double-Mean method takes into account the experimental uncertainties of different authors and all the available experimental information is fully considered. Thus the evaluated values of DM are less dependent on discrepant data and the uncertainties of the evaluated values are more reliable, i.e., the deviation between the evaluated results and the "true" values smaller than the other evaluation methods. The application of using the measured half-life and γ emission probability for ⁷Be is given as an example. The results of half-life and γ emission probability deduced from the present work are T_1/2=(53.282±0.012)d and P_γ=(10.45±0.04)%, respectively.