(more see downloads)

Data evaluation in complex issues like human factors often means dealing with multiple relationsships between the parameters that are investigated. A classical example of such interrelations is the one between workplace layout and salary with individual motivation for instance.

On the other hand, extensive investigations are needed for getting data about such relationships on a statistical sufficient basis that a factor- or cluster-analysis would require. Often such extensive empirical data exemption is neither economic, nor feasable nor does it bring any advantage in term of precision or relevance of the statements derived from the study.

NMDS (Non Metrical Multidimensional Scaling) is a tool to perform quasi-cluster analyses or factor-analyses with data that cannot be evaluated statistically by cluster analyses or factor-analyses. The following picture shows an example of influencing factors on human reliability from CAHR:

NMDS methods are best fit iterative methods to find the relationships between parameters. One problem of NMDS methods is that they tend to fall into local minima, i.e. they do not find the best fit in a set of parameter.

The program NMDS avoids this by using a temperature function. Temperature functions are well known means in AI (Artificial Intelligence) to avoid that neural networks fall into local minima.