A test of measurement results with increased sensitivity to inconsistency

Abstract This paper demonstrates that the chi-square test of consistency of a set of measurement results is improved upon when each measured value is compared with the weighted mean of all the other measured values, not the overall weighted mean. The corresponding test is a Monte Carlo test, where the null distribution for the new test statistic is obtained by simulation. Basing the procedure on the measurement results with smaller uncertainties would give further improvement.