Points of view: Power of the plane

readily explored using a grid of scatter plots that represent all pairwise combinations. The choice between a parallel coordinate plot and a scatter plot matrix depends on the analytical task to be supported. The fundamental difference in the approaches is how they represent individual data features across multiple dimensions. A data point in a parallel coordinate plot is depicted as a single line or a profile (Fig. 1a,b). Together, the ‘bundles’ of lines point out clusters, and outliers therefore become apparent. A scatter plot matrix, on the other hand, represents a data feature as a series of points that are not connected across the scatter plots, making it difficult to draw conclusions about individual data features (Fig. 1c). However, scatter plot matrices can be used to efficiently identify pairwise correlations and other relationships between all dimensions in the overall dataset based on the characteristic shapes of the point clouds. These methods complement each other and will deliver the best results when used in an interactive setting in which multiple coordinated visualizations of the same data set are available. Along with heat maps and dimensionality reduction tools, fundamental 2D visualization methods can be powerful approaches to multivariate data. ComPetinG FinanCiaL interests The authors declare no competing financial interests. nils Gehlenborg & Bang Wong