Extreme Outlier

Purpose

Extreme Outlier is a statistical computer program eliminating extreme outlier values from statistical probe series.
Detailed description of the extreme outlier method can be find in the article


Jack Shoemaker, Robust Outlier Identification using SAS


The author's original implementation of the method uses licensed software from the SAS Inc. company. Our implementation is developed in Java language and uses free Java libraries.

How it works

Extreme Outlier progam can be started in the command shell environment in many operating systems ( Windows, Linux, Unix, Mac ).
Extreme Outlier processes data delivered in simply formatted files: the probe data are formatted as numbers  ( with comma as decimal separator ), separated by newlines.
Such a file can be simply created by using simple text editor or exported to CSV format from Excel spreadsheet.

As a result of processing and analyzing input data, Extreme Outlier produces text report file, documenting probe series, exposing the extreme outlier values and additionally reporting certain statistical values:

Z Value of 25th percentile
Z Value of 75th percentile
Inter quartile range
Lower bound
Upper bound
Area under lower tail
Area under upper tail
Two sided alpha value

The Extreme Outlier program is started by calling the command:

java -cp commons-math-1.2.jar;outlier.jar;. com.mpsystem.stat.outlier.OutlierFilter sourcefile reportfile s 1,5 0,0005

with parameters representing

-cp commons-math-1.2.jar;outlier.jar;.      - the classpath of the required libraries

com.mpsystem.stat.outlier.OutlierFilter    - the called program class main method

sourcefile   - source file with input probe data

reportfile - output report file

s - single run ( input data representing single course of probes in single file )  or m - multiple run ( input probes in multiple input files representing multiple series )

1,5 - multiplication factor ( parameter of the extreme outlier algorithm: multiplication coefficient influencing the breadth of the accepted values band )

0,0005 - significance ( parameter of the extreme outlier algorithm: coefficient specyfing the significance of rejected values  )

 

Input data must be supplied as a file of newline separated numbers.

 

How it can be used

Extreme Outliner is completely developed in Java programming language and therefore it can be used ( after having installed the Java runtime -> www.javasoft.com ) on any operating system !

Installation

Here can you download runtime version of our program.

Requirements:

1. JRE/JDK - Java runtime environment ( to download goto www.javasoft.com ) in version >= 1.5

2. Free statistical library commons-math ( to download goto http://commons.apache.org/math ) in version >= 1.2

3. JAR file of the Extreme Outlier program

4. Start file outlier.bat, using additional batch file cpappend.bat.

To download our Extreme Outlier program click on:

outlier.zip

If you are interested in obtaining the source code and support, please contact us directly.

Licence

Extreme Outlier is a product of a MP System Spzoo company. All rights reserved by MP System Spzoo. MP System grants to all private, commercial or institutional users unlimited usage no-fee licence. Responsibility  for any damages resulting from using this product is excluded.