Abstract
Data visualization – that is, the graphical representation of numerical information – is foundational to the scientific enterprise. A broad literature base is available providing rules, guidelines, and heuristics for authors of scientific literature to assist in the production of scientific graphics that are readable and intuitive. However, most of the available recent publications are in the bio-, psycho-, or climate sciences literature. In this paper, we address this deficiency and provide data visualization heuristics tuned to the specific needs of the physical sciences, and particularly materials sciences, community. We enumerate six general rules and provide examples of bad and improved data graphics, and provide source code to illustrate the generation of the improved figures. The six rules we enumerate are: (1) Generate figures programmatically; (2) Multivariate data calls for multivariate representation; (3) Showing the data beats mean ± standard deviation; (4) Choose colormaps that match the nature of the data; (5) Use small multiples; and (6) Don't use vendor exports naïvely.
