- Norman Mailer, upon being told to stick to the facts

during the trial of Chicago 7

Naturalistic University of Alberta Non-linear Correlation Explorer (NUANCE) is a computer program that can be used for data exploration and analysis. NUANCE searches for variable relations by evolving mathematical functions though natural selection to maximize the correlation between the functions' output and a dependent measure you specify for NUANCE. NUANCE is written in the Java programming language, and is compatible on all common operating systems.

1. What does NUANCE do?

3. Why should I use NUANCE?

4. Where can I get NUANCE?

5. How do I use NUANCE?

6. Reporting a bug

7. Can I contribute to the project?

Nuance uses the Genetic Programming paradigm to create and breed mathematical formulae. It compares the output of each formula against 'best fit' values (the values of a dependent measure you specify for the program), and then ranks the formulae from best to worst in predictive power. The worst formulae are killed off, and the best mate with each other to produce offspring, which are blends of two parent formulae. The process is repeated for many generations, each usually increasing the predictive power of the best formula.

NUANCE is an excellent tool for data exploration. NUANCE does not require any a priori assumptions about your data, and it has the power to capture nonlinear variable relationships, which is hard to do with common statistical techniques. It is fast, efficient, easy to use, and it provides a very explicit mathematical summary of the datasets you provide it with. Furthermore, the mathematical nature of NUANCE's output makes it quite easy to extract information from NUANCE. If you would like to understand the relationships of a set of variables, NUANCE is an indispensable tool.

The latest version of NUANCE can always be found right here. The latest official release of NUANCE is v2.0, which was released Friday January 7, 2005. However, we are currently working on a new version of NUANCE already with substantial improvements over v2.0, both in terms of mechanics and user interface. In addition, the new version will be open source. Thus, if you are interested in NUANCE's source code, it is suggested you download this version for the time being.

A documentation should be included in the nuance.zip file here. Alternatively, you can download just the documentation separately from here.

Though NUANCE has been checked for bugs quite carefully, it undoubtedly still contains errors. If you have found a problem with the program, please email a description of it to Geoff Hollis. A fix will be incorporated into the next version of the program.

If you have a suggestion about how to improve the program, feel free to email Geoff Hollis or Chris Westbury with a description of your idea. Another way to contribute to the project - and one that we request if you use NUANCE - is to cite the program for any scientific work you do with it. The proper APA citation form is:

Hollis, G. & Westbury, C. (2003). NUANCE: Naturalistic University of Alberta Non-linear

Correlation Explorer. http://www.ualberta.ca/~hollis/nuance.html