Authors: Nils Rodrigues, Daniel Weiskopf
Abstract: Conventional dot plots use a constant dot size and are typically applied to show the frequency distribution of small data sets. \ Unfortunately, they are not designed for a high dynamic range of frequencies. \ We address this problem by introducing nonlinear dot plots. Adopting the idea of nonlinear scaling from logarithmic bar charts, our plots allow for dots of varying size so that columns with a large number of samples are reduced in height. For the construction of these diagrams, we introduce an efficient two-way sweep algorithm that leads to a dense and symmetrical layout. \ We compensate aliasing artifacts at high dot densities by a specifically designed low-pass filtering method. Examples of nonlinear dot plots are compared to conventional dot plots as well as linear and logarithmic histograms. \ Finally, we include feedback from an expert review.