Advanced charts

Alluvial, or Sankey diagrams, are a type flow diagram designed to show two indicators of a dataset and how records distribute among them, highlighting correlations.

Multiple blocks of two indicators can be placed next to each other creating a wider alluvial diagram but it’s crucial to consider that this specific kind of chart does not show correlations between indicators that are not directly connected. This is usually emphasized using different set of colors for each block.

A heat map is a graphical representation of data where the individual values contained in a matrix are represented as colors. When it comes to heat maps, the most common color ranges are expressed in either sequential and diverging color scales according to the typology of data represented.

Sequential scales use a blended progression, typically of a single color, from the lightest to the darkest values, representing lows to highs. It’s best practice to use a sequential scale with values that are all positive or all negative. Sequential 
heat maps can leverage the full range of the palette (from 10-100) to maximize contrast. Please note that the 3:1 minimum contrast requirement does not apply to heat maps.

Axis ticks and legend can be used to leverage the chart potential giving the users additional information and dedicated kinds of interaction.

Diverging scales show color progression in two directions: dialing down the first value from one end to a neutral color at the inflection point (for example, a 0 value midway between -100 and +100), then increasing the opacity of the second value to the darker end of the scale.

Divergent scales require 3 colors for their 3 correspondent values (min, max and inflection point) and they are referred to as “symmetrical” when the min/max values match and “assymetrical” otherwise. Although some examples use the stepped scale provided by the Carbon palette, divergent heat maps often leverage continous values to accomodate situations that require more than 10 values.

While min/max values can be automatically detected from data, inflection points can vary depending on data type. We default to zero in order to accomodate the most common cases and include a feature flag within the code to specify exceptions.

A network diagram is a way of visually representing network architecture. It maps out network structure with a variety of icons and connecting lines. It’s ideal for sharing the layout of a network because the visual presentation makes it easier for users to understand complex connections at a glance. These diagrams are especially useful for isolating problems or when designing a new system.

The example below illustrates a swimlane graph with expandable nodes developed by the Security team. In products like QRadar these diagrams are often paired with side and filter panels. However, a simplified version of the component is available in the Carbon Add-ons repo for IBM designers to adapt.

This specific network diagram is comprised of parent and child nodes. In the QRadar example, parent nodes convey the story of what happened in an investigation and can be clustered based on relationship type. Child nodes appear under parent nodes and can also be clustered.

Network diagrams use the same scale controls that apply to the rest of the data visualization components, however they appear in a floating tile, enabling the user to zoom and pan simultaneously. It’s also helpful to include the expand and collapse all nodes feature, which gives users the ability to easily navigate between detailed and high-level views of the diagram.

Parallel coordinate charts visualize individual data elements across multiple variables. Each of variable corresponds to a vertical axis and each data element is displayed as a series of connected points along the axes.

The parallel coorinate chart is the equivalent of a radar chart with axes — making it preferable for many use cases. Its strength is that each variable can be completely different and even have its own scale (the units can even be different).

Hover effects that highlight a specific group or axis can be added to enhance basic parallel coordinate charts. These charts can also include interactive features like the brush component. By clicking and dragging along any axis, you can specify a filter for that dimension. The brush component is also used in the updated scatterplot matrix example. See the detail below.

Treemaps can work for visualizing a part-to-whole relationship among a large number of categories—as long as the data is hierarchical and exact comparisons between the categories is not a primary concern.

When dealing with large amounts of data in a constrained space, tree structured nodelink diagrams (like the network diagram shown above), grow too complex to be useful. Tree maps provide an attractive alternative by maintaining a hierarchical structure while displaying rectangular quantities for each category and subcategory via area size.

“Trheat” maps combine the divergent color scale of a heat map and a tree map. The “trheat” map is identical in terms of structure to the basic tree map, but the countries can be colored independently accordingly to an additional indicator or variation of an indicator. Again, the diverging palette shows color progression in either direction from an inflection point (for example, a 0 value midway between -100 and +100),

A map that uses differences in shading, coloring, or the placing of symbols within predefined areas to indicate the average values of a property or quantity in those areas.

Symbols driven by data are overlayed in geographical regions. A bubble the symbol most commonly used in this instance—with the area of the circle proportional to its value in the dataset.

Connection Maps are drawn by connecting points placed on a map by straight or curved lines.

While Connection Maps are great for showing connections and relationships geographically, they can also be used to display map routes through a single chain of links. Connection Maps can also be useful in revealing spatial patterns through the distribution of connections or by how concentrated connections are on a map.

Density-equalizing cartograms are the most common type of cartogram. In this situation the mapping variable takes the place of the land area or distance in the map, causing the map to become distorted in proportion to the substitute variable.

Cartograms are useful for visualizing populations through different lenses — for instance, ethnicities, polictical parties or religious affliliation. 

Carbon does not have a coded component for cartograms but online generators like go-cart.io allow you to upload your own data sets and download an .svg file.

The Dorling Cartogram is a technique for representing data for areas that avoids geography in favor of (normally) a geometric shape that represents the unit areas. Circles are usually chosen since they can be neatly positioned.

As with the density-equalizing cartogram above, the larger the total population, the larger the representation of the country.

For more dynamic, real-time mapping, Carbon has created four themes using Mapbox. Mapbox is a developer platform used across industries to create custom maps that can be easily integrated into websites and apps. These themes can be used as is, or in conjunction with Carbon-styled data points.

These styles can be used as is, or as a backdrop for other geospatial data sets. For more complex data sets that require the full breadth of the Carbon charts palette, we suggest using either the White or the Gray 100 style for the most accessibile results.