Silva, Renato R. O. daRauber, Paulo E.Martins, Rafael M.Minghim, RosaneTelea, Alexandru C.E. Bertini and J. C. Roberts2015-05-242015-05-242015https://doi.org/10.2312/eurova.20151100Multidimensional projections (MPs) are key tools for the analysis of multidimensional data. MPs reduce data dimensionality while keeping the original distance structure in the low-dimensional output space, typically shown by a 2D scatterplot. While MP techniques grow more precise and scalable, they still do not show how the original dimensions (attributes) influence the projection's layout. In other words, MPs show which points are similar, but not why. We propose a visual approach to describe which dimensions contribute mostly to similarity relationships over the projection, thus explain the projection's layout. For this, we rank dimensions by increasing variance over each point-neighborhood, and propose a visual encoding to show the least-varying dimensions over each neighborhood. We demonstrate our technique with both synthetic and real-world datasets.H.5.2 [Information Interfaces and Presentation]User Interfaces10.2312/eurova.2015110031-35