I am happy to announce Sliver, a free software application I wrote over the last two years for multivariate data visualization. Sliver includes parallel coordinate (PC) plots, PC plot matrices, various types of configurable 2D and 3D scatterplots, and plots overlaid on Google Earth, all fully linked by color brushing. Transparency (alpha blending) is supported as seen at left. Sliver also offers data animation, including Google Earth animation as well as the Grand Tour, a rotation through n-space that reveals correlations and structures in multidimensional data. Have a look!
This modern application may seem at odds with the theme of my blog, but in fact it was Maurice d’Ocagne who coined the term parallel coordinates in regard to his parallel scale nomograms. A hundred years later Alfred Inselberg extended this idea of parallel scales as a way of visually analyzing multi-dimensional data. Inselberg and others make use of some of d’Ocagne’s work on point-line duality to characterize functional relationships between variables as structures and envelopes of the lines drawn between their axes.
In a parallel coordinate plot the axes of values for the variables lie parallel to each other, typically running vertically with a linear mapping from the minimum to maximum value of each variable. Each instance of measurement of the variables (i.e., each row of the input CSV data file) is represented by a segmented line, or polyline, that passes through the corresponding value on each variable axis. Correlated groups of lines can be color brushed, and this color brushing propagates across any other 2D or 3D plots generated from the data. The parallel coordinate plot shown here does not display axes or labels, but this larger version does display axes labels (best viewed by downloading and opening it rather than in your browser).
These highly visual displays reveal correlations, patterns, trends and anomalies in multivariate systems, and these in turn can significantly aid in the diagnostic analysis of sensitivities and error sources in a system. I have been an advocate of nomograms as system models, and this aspect of parallel coordinate plots is what drew me to study them and ultimately to create Sliver. This linkage will be more complete when I report later on the use of Sliver to test an application of nomograms to image processing.