Table 1.
The levels are compared using 4 measurement-related classes: resolution, property, mathematical operators, and central tendency [2].
Table 2.
Pros and cons of commonly used color spaces.
Fig 1.
Example of 3 color harmonies in the key of cyan.
These harmonies were created using the Adobe Color web tool (color.adobe.com). They are color blind friendly palettes and are presented in Web Hex format. Monochromatic: 2C7C9D, 65BFDA, 39484C. Analogous: 5FE896, 5FF3E3, 3CA7D2, 1E78EF, 1938E3. Complementary: 22607C, 3CA6D0, 4CCFFA, D06D21, 7B3514.
Fig 2.
Example of 3 color palettes according to the 3 main data types.
These color palettes are based on the ColorBrewer schemes. They comprise different classes and are given specific names in ColorBrewer: sequential (9-class Blues), diverging (11-class RdBu), and qualitative (12-class Paired).
Fig 3.
Example of heatmap color mappings.
Left: Bad example where a diverging color palette (YlGnBu) is applied to ordered data that progresses from low to high (0 to 1). The color mapping represents higher values in lighter colors. Right: Better example where a sequential color palette (Purples) is applied to the data. The color mapping reverses the importance. Gray cells depict missing data.
Fig 4.
Example of data visualization with a color context-related problem.
(a) The example is shown on a computer screen with a gray background versus printed on a paper or shown on a white background. (b) Alternative color encoding of the line using a yellow color solving the problem.
Fig 5.
Example of correlation matrix plot with upper triangle.
Color intensity and the size of the circle are proportional to the correlation coefficients. Left: Chromatic aberration with the red/blue color combination. Negative/positive correlations: red/blue. Right: Improved data vis with the green/purple complementary color combination. Negative/positive correlations: green/purple.
Fig 6.
Nonuniform distances between hues in rainbow-based color palettes.
(a) Typical rainbow colormap used in visualization tools and analyses. (b) Nonuniform distances between hues. Image adapted from Applying Color Theory to Digital Media and Visualization, p. 34, Problems with the Rainbow Colormap [12].
Fig 7.
Penicillin and neomycin resistance of bacterial strains.
The biological data vis is colorized based on the nominal variable: Gram staining. Left: Domain-independent colors. Right: Domain-dependent colors. These colors better reflect the actual gram staining colors seen under a microscope. The reverse log scale of the MIC is shown for each axis. MIC, minimum inhibitory concentration.
Fig 8.
Example test of dichromatic views using Coblis for data vis examples created by Viz Palette.
The color palette was created for 4 data classes using ColorBrewer by selecting the qualitative scheme and the color blind safe option. Although the palette is color blind safe, individuals with the very rare tritanopia will have a hard time distinguishing the classes.
Fig 9.
Example of heatmaps adapted from Fig 1.
Left: Heatmap obtained by converting Fig 1 (right) to gray scale. If the only relevant information concerns the min and max values, this heatmap is suitable. Right: Midrange values are more visible, thanks to a negatively clipped sequential color palette, i.e., the mapping of the data values is shifted to the darker range of the gray scale. The values of the Jaccard index around 0.5 are brought into the foreground and are more visually pronounced.
Fig 10.
Example of upper triangle adapted from the correlation matrix in Fig 2.
Only the size of the circle is proportional to the correlation coefficients. The background is shown in light gray; otherwise, white circles are not visible. Positive/negative correlations are color coded, either in black/white (left) or in white/black (right). Depending on the story we want to convey, the visual importance of positive or negative correlations can be emphasized using black. To limit the focus on 1 specific range of the data, it is also possible to either color code 1 correlation type using a gray scale palette. Since the color of some gray circles may coincide with the background color, one should be aware of such an influence on the audience's perception.
Table 3.
Summary table describing the purpose of each rule.