Circular background decreases misunderstanding of multidimensional scaling results for naive readers

Abstract

Non-linear multidimensional scaling (NL-MDS) methods are widely used to give an insight on structures of a dataset. Such a technic displays a “map” of data points onto a 2 dimensional space. The reader is expected to have natural understanding of proximity relationships between items. In our experience, MDS are especially helpful as a support for the collaboration between data analysts and specialists of other fields. Indeed, it often allows understanding main issues, major features, how to deal with data and so on. However, we observed that the classical/rectangular display of map causes confusion for non-specialists and long explanation is often required before reaching the fruitful step of the collaboration. The meaning –the absence of meaning, actually- of axes can be subject for many questions and skepticism from many naive persons. Although it is hardly quantifiable, we observed that using a circle-shaped background for maps improves the understanding of the concept of data mapping by far. We however present here a subjective feedback that may support the practical contribution of NL-MDS for other scientific fields.