by Robert Simmon
The use of color to display data is a solved problem, right? Just pick a palette from a drop-down menu (probably either a grayscale ramp or a rainbow), set start and end points, press “apply,” and you’re done. Although we all know it’s not that simple, that’s often how colors are chosen in the real world. As a result, many visualizations fail to represent the underlying data as well as they could.
Read the blog series and optionally also watch the lecture.
Sequential datasets are best represented from with simoultanious shifts in hue, saturation, and lightness. Best used to depict data that varies continuously from high to low.
Divergent datasets are best represented as a smooth continuous linear palettte that use variations of hue, lightness, and saturation. It is best to use two different, contrasting, hues with a neutral central color (white) producing a centralized white. This allows our eyes to quickly diffrentiate data.
Qualitative data sets are best to be seperated into groups of different distinguishable colors (idealy less than 12). With increased data, more groups need to be made with additional lables and symbols when needed.
Good use of different color palettes can enhance the storytelling, clarity, and attention aspects of different datasets. I think the blog series introduced the ideas of applying color scales to data nicely. It discussed about three types of datasets: sequential, divergent, and qualitative. These theories are interesting to me. I do know color is a significant variable in datasets, but never thought too much about how different color schemes are suitable for different datasets.
Sequential datasets are suitable for colors scales that are smooth and even. The lightness and saturation of the palette should be uniform. Since this type of datasets includes continuously varying data (ex: from low to high), the color scheme should also reflect the continuously changing quality.
Divergent datasets focus on measuring differences. The concept of divergent palettes is introduced, which is composed of two sequential palettes merging with a neutral central color. This color is usually white and gray.
Qualitative datasets categorize things. As data is separated into different categories and classes, the colors for qualitative datasets should be as diverse and distinct from each other as possible. Because people all perceive things differently, and also because of the simultaneous contrast effect, the maximum number of categories that can be displayed and recognized is probably no more than 12.
I also think it is a big challenge to design a graph with color that can be perceived by normal people and colorblind people at the same time. What will a qualitative dataset look like under this idea?
The three different types of datasets discussed are sequential, divergent, and qualitative. The series of posts does a great job with analyzing the correct color palettes to use in each situation. When used correctly, color has the ability to enhance quality, aid storytelling, and draw a viewer in. How does color contribute to multi-level storytelling? In what ways can color create hierarchy? I thought it was most interesting to see how color placement can evoke a sense of depth. If used incorrectly, this sense of depth can skew data and create misleading information. Therefore, color truly does play an important part in making data easily accessible while maintaining the data’s integrity.
Sequential datasets require simultaneous shifts in hue, saturation, and lightness because they vary continuously from high to low.
Divergent datasets make use of continuous linear color palettes that rely on variations of hue, lightness, and saturation. Two contrasting hues with a neutral middle color is the best for this dataset.
Qualitative data are best to be represented by various colors since the increased amount of data requires a higher level of differentiation.
After reading the series of articles I am left with questions pertaining to how we can make color accessible for blind people? Especially since color is an integral component to understanding the information presented. Also, I wonder what determines if use of a color is qualitative or quantitative. Qualitative data uses color to separate areas into categories. Therefore, it is is best to have very different colors in the set to avoid confusion. However, I am curious how colors that are too similar can possibly affect the integrity of the data. It seems that color choice is important in providing a viewer with clarity.
Subtleties of Color presents the dilemmas and their corresponding solutions of color choices in data visualization. Since, unlike computers, the human eyes perceive colors subjectively, color design precautions are necessary for a data visualization to serve its purpose of illuminating data.
Based on the three different types of data––sequential, divergent, and qualitative, the three basic aspects of colors, hue, saturation, and lightness, are to be manipulated separately or collectively to fit the dataset. Sequential data is best displayed with a palette that varies in both lightness and saturation, with the supplementary addition of shift in hue; divergent data is best displayed with bifurcated palettes with a neutral central color; qualitative data should be represented by a set of easily distinguishable colors, or colors that vary sharply in hue, saturation, or lightness. In general, palettes appropriate for its content and distinguishable to human perceptions are the perfect palettes.
In addition to the fundamentals, there are some subtle aspects of color design for data visualization. Considering the perceptual norms of colors, the color palettes should be tailored for each dataset based on intuitions, layering, complementary datasets, breakpoints, separations, and hierarchies.
Specifically noting the perception of color, I am wondering whether or not certain ranges of colors would not be perceived as different from one another. Not to the extent of color blindness, in which green and red are highly indistinguishable, but rather in the case of the equiluminant colors chart where the spread is radial. The closer to the center of the circle the colors get, the more vague the color differences end up being. In that sense, we can see the same effect happen in our Retinal Variables charts, for the hue and value gradiations the initial start was nearly unperceivable.
In that sense, it seems to be necessary to avoid the usage of green, as the general blend of color seems to accentuate this problem even more so than other colors. Color palettes that have both a progression in lightness as well as a shift in hue seem to be the best solution as the two ends of the color spectrum, as a result, vary so drastically from one another that the middle increments have a more distinguishable step up from the previous.
On the otherhand, it is interesting that for Divergent Data, because the focus is more on the two ends rather than all values, the full palette tends to want more of the eye catching variances to avoid the center range. Categorical Data focuses on distinct hues for each value, with the reading stating that the maximum number is around 12, or possibly even fewer. To go over this estimation would mean that certain colors may start to resemble another specific color on the spectrum, which may cause confusion in perceiving the chart if it is a large scatter of colors.
In this reading, Robert Simmon details the subtleties of color, and notes three types of datasets, those being sequential, divergent, and qualitative.
Forgetting the specific definition of these datasets for a second, we first explore the fascinating ideas behind color and the power it has on perception and audiences. Different colors can enhance storytelling, graphics and work to make the viewer more interested and more engaged with the content in front of them. Simmon works to explain and analyze how color palettes should be shifted situationally, playing with ideas of hierarchy, depth and data visualization.
Regarding the three types of datasets, the first is sequential datasets, which are represented with shifts in hue, saturation, and lightness simultaneously. This type of representation is most apt in order to depict data that has variations continuously from high to low.
Divergent datasets are useful in order to focus on differences. The color palettes for this type of dataset rely on variations of hue, lightness and saturation, and work the best when two contrasting hues and a neutral middle color are used.
Qualitative datasets are used to categorize items or data. Similar to how the data is separated into different categories, the use of various colors for qualitative datasets mirror the differentiation of categories and data.
I must note that while I understand how color can affect mood, tone etc, and have aptly used it in other design and artistic endeavors in order to convey the feelings I want, I had never consciously considered how utilizing varieties of color schemes and palettes would be more apt depending on the data. Rather, I believe I was more concerned with color in order to simply provide clarity and ease of understanding, and thinking about color palettes subconsciously.
It is known that humans are unable to process all colors of the world. With our limited perceptions of color, the tools we use to transmit color data become crucial in our daily life. Color profiles like RGB, HSB and CIE L*C*h served different purposes depending on the ultimate medium. The fact that humans are less able to perceive saturated blue than red and green has to be taken into when designing data visualizations.
The method of increasing lightness while shifting hue is a wonderful solution for visualizing the scale of data. However it is only appropriate for sequential datasets. It is also interesting to see the colors used in divergent datas and the importance of catering towards color blind individuals. Categorical data on the other hand is tricky to manage as a very controlled color palette is needed.