Abstract

Perceptual mapping is widely spread to assess perception in different areas, like marketing, political and social sciences, psychology and others. One opportunity for development is to add statistical inference on final configuration in order to consider inherent differences of a group of evaluators. The main objective is to produce multiple perceptual maps from focal panel and to incorporate the confidence regions of different evaluators into the visual representation using MDSvarext. The algorithm represents a joining of non metric multidimensional scaling, shape statistical tool, clustering techniques and non parametric estimation of variance-covariance matrix to generate a visual representation of object's perception and its confidence regions. An experiment to assess occupational risk perception has been run in order to demonstrate the method. The results showed that different perceptual maps are needed to encompass the variability of a focal group. The generated perceptual maps have different interpretations since the objects may be on opposite sides of the graph. The solution generated by MDsvarext was effective and statistical inference could be done. To explore the variability in focal groups is very important, and MDSvarext represents a path to be followed, since it was possible to visualize the differences that are statistical significant.