Volume 28, Issue 2 (March 2000)
The U Test for Outlier Detection in Normal Distributions with Known σ
A new criterion is proposed for testing for outliers when the population standard deviation a is known. The advantage of this criterion is that no table is required and a fixed significance level can be selected. Compared with the current ASTM E 178 criterion for outlier rejection, under the same chance of making a Type I error, the U test has less chance of making a Type II error and therefore is a more powerful test.