Discordance Outlier Test

 

See Also:

Dixon's Test for Outliers

Rosner's Test for Outliers

 

Description:

 

The Discordance Outlier test is a relatively simple outlier test presented in USEPA (2000). The test can only determine if the highest value in a data set is an outlier. After removal of the outlier, data must be normally distributed (or transformed to a normal distribution) for the test to be appropriate.

 

 

Remarks:

 

At least 3 measurements are required for this test. Lookup tables provide critical values for up to 50 measurements. For data sets with greater than 50 measurements, the critical value for 50 measurements is used as a close approximation.

 

ChemStat can perform this test at either the 99% or 95% confidence levels.