Rosner's Test for Outliers

See Also:

Description:

Rosner's test is a procedure for detecting up to 10 outliers in data sets with 25 or more measurements. The data (or transformed data) must follow a normal distribution. A strength of the procedure is that it detects outliers that may be masked by other outliers.

The ChemStat implementation follows the procedure described by Gilbert (1987). For the selected parameter, the test can be performed on samples from a single well, all compliance well, all background wells, or all wells. The test can be performed at either the 5% or 1% levels of significance.

Rosner's test is two-tailed meaning it will detect either suspiciously large or suspiciously small data.

Use:

For detecting up to 10 outliers in normally or log-normally distributed data sets with 25 or more measurements.