What do we mean when we talk about details-on-demand?
Exploring data to find insights usually follows a familiar process. We clean it up, create visualizations, interact with them, do some analysis. Rinse and repeat. We don't usually have to think much about the process, which is a testament to how easy it is to do it in JMP.
But there is an underlying process to exploratory data analysis. It has been the subject of many studies over the years. One of most influential is "The Eyes Have It: A Task by Data Type Taxonomy for Information Visualizations" (Shneiderman 1996). This seminal paper introduced the Visual Information Seeking Mantra, or Infovis Mantra for short. It goes like this:
Overview first, zoom and filter, then details-on-demand.
After we create a visualization, more often than not, we have to interact with it to find interesting patterns. Zoom and filter change the visualization focus. Details-on-demand add extra information (details) only when it becomes necessary (on demand). To reduce the impact on the exploratory flow, we use the mouse to point to the area of interest. A brief pause triggers the interaction. And voilà: a hover label pops up showing details, on demand.
How do we decide which extra information is relevant in this context? First, JMP gathers data from Labeled Columns, or row numbers if none is defined. Then it adds information extracted from the underlying visualization. For example, the numeric or categorical values associated with the underlying visual element. These can be extracted from the X and Y axis. The final result is displayed as text in the hover label window. The window can then be pinned to the graph, repositioned, or dismissed by moving the mouse away.
Beyond Text
JMP 15 introduced hover label extensions. These features expand the type and content that hover labels can display.
- Textlets allow the display of entire paragraphs of rich text. The content can be static or dynamic, based on the evaluation of JSL variables.
- Gridlets provide fine control over the hover label default content, or grid. JMP builds this content from labelled columns and graph roles. The gridlet extension can rename, remove, reformat and restyle grid entries. It can also add new ones, with both static and dynamic content - including clickable links.
- Graphlets display graphical or visual content inside the hover label window. The most common usage is to show new visualizations based on the hovered visual element data. For example, a distribution of the values aggregated by the bar in a bar chart. Other possibilities include previewing associated image files and even web content.
Here is a quick example showing the basic hover label details-on-demand interaction, and how it can be extended using graphlet presets.
Beyond Details
The Infovis Mantra is comprised of four tasks, details-on-demand being the last one. But Shneiderman's paper describes three more, which the new hover label extensions support.
- Relate: The Paste Graphlet creates thumbnails from preconfigured platform graphs. Use it with the Fit X by Y platform to visualize correlations in the current data context. Other platforms can be used to show different types of relationships.
- History: The graphlet thumbnails are clickable images. When clicked, they launch the platforms used to generate them in independent windows, or they can launch in place, replacing the current visualization. In either case, an Undo action recovers the original visualization. You can drill-down many levels knowing that you can always go back.
- Extract: Visual elements correspond to subsets of the data. The hover label menu gives direct access to these subsets. Also, launched graphlets include a pre-configured Local Data Filter, which allows the associated query to be copied or even modified.
More documentation on hover labels is available at the JMP website.
Beyond This Post
In the next posts, we will cover different aspects of the hover label extensions architecture. The first one talk about the high-level interface that exposes the vast majority of the new functionality. No coding skills required!
In the meantime, check our Discovery 2020 presentation about this same topic: From Details-on-Demand to Wandering Workflows: Getting to Know JMP Hover Label Extensions
References
[Shneiderman 1996] "The Eyes Have It: A Task by Data Type Taxonomy for Information Visualizations," by Ben Shneiderman. Department of Computer Science, Human-Computer Interaction Laboratory, and Institute for Systems Research. University of Maryland College Park.
Further Reading
B. Craft and P. Cairns, "Beyond guidelines: what can we learn from the visual information seeking mantra?," Ninth International Conference on Information Visualisation (IV'05), London, UK, 2005, pp. 110-118, doi: 10.1109/IV.2005.28.
Now You See It: Simple Visualization Techniques for Quantitative Analysis
Stephen Few, 2009
Improve Your Vision and Expand Your Mind
Stephen Few, 2007
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.