Here is another form of diagrams, that can not be built in Excel. Generally, in most cases, you can live without it, but there are times when it can be invaluable.
Imagine: you want to compare 20 different cars on a number of factors: acceleration to 60 mph, horsepower, swept volume, buyers evaluation, and etc. All these criteria have different scales and units: seconds, horsepower, subjective evaluation from 1 to 5, liters … In such cases, there are two standard solutions:
- Draw a separate graph for each indicator; in this case you will not be able to compare one car with the other under all characteristics.
- Compose a table that takes a lot of time and attention to read
As a helpful solution, here is parallel coordinates where all scales are normalized:
Example above illustrates social-demographic analysis. In contrast to Excel, where we can fit only 3 axis (one on X and two on Y), there are no restrictions.
If there are less than 5-7 axis, they can be composed to make up a circle; similar meaning axis can be placed closer to each other. The example below is taken from www.Hiveplot.com — where you can find many interesting things. Its author Martin Krzywinski describes the possibility of using these diagrams in science (particularly in description of human genome). He calls these graphs — Hive graphs, and considers them applicable only for large data sets.
Mathematical reasoning of these graphs is described in a book called «Parallel Coordinates.Visual Multidimensional Geometry and Its Applications.» First 50 pages are available here.
- There is a link to the tool on Hive Plot http://wodaklab.org/hivegraph/graph/index, but it looks a little too raw
- I also found a software for interactive visualization — Ggobi, and a reference to the parallel coordinates — http://www.ggobi.org/docs/parallel-coordinates/
- Infoscope — be sure to review
- Last but not least: recently, our beloved information design agency, Density Design, has announced the release of Fineo. They went beyond the previous resources and added thick lines, thus combining the parallel coordinates with Sankey. In addition, there is an option of building vertical and horizontal models. It looks just fine.
The only disadvantage is the format in which data is fed to the system. It’s a text file created by certain rules — the template can be downloaded here. But it didn’t help me much:) Nevertheless, we wish them good luck in automation of visualization tasks!