Convergence of techniques for the evaluation of discrepant data.

The problem of evaluating discrepant data has been addressed by several authors over the previous 20 yr. More recently some attention has been given to the use of the median, which is expected to have better statistical 'robustness'. The various evaluation techniques should converge towards the 'true' value as the number of data in a data set increases, and the 'robustness' of each evaluation technique can then be tested by the rate at which that technique converges. Several evaluation techniques have been applied to discrepant data sets, and the results are shown to converge as the size of the data set grows. The discrepant data sets used as examples are the measured half-lives of 90Sr and 137Cs. Differences in the behaviour of the evaluation techniques are discussed, as applied to these data sets. The half-lives deduced from this study are: 90Sr 10,551+/-14 days; 137Cs 10,981+/-11 days.