Filtering those spaghetti in bump chart

Bump chart is a two-dimensional ranking visualization tool. It allows us to visualize data rank by one dimension which changes along the other dimension. One example is the soccer league ranking by team along weekly schedule. Here is a video that teaches us how to create a bump chart:
http://www.tableausoftware.com/learn/tutorials/on-demand/bump-charts-0

The Limitations of Bump Chart

Bump chart has its own limitations: When there are a few teams, it is easy to see and track the rank changes. However, when there are a lot of teams, the chart tends to look like spaghetti, or a piece of modern art. Sometimes I would like to view only one spaghetto, that is, one team's ranking along the weeks. This blog will provide a solution for doing just that. Hopefully, the technique presented here will apply to using bump charts in cases of ranking large dimensions.

The following is a bump chart that ranks 51 states (including DC) per each major cause of death. There are 24 major causes of death in total being considered. It seems too much for bump chart to handle.

We can somehow use the color legend to highlight each state's ranking. But the long list of states can't be minimized as a drop down menu and look quite cumbersome. So it is not ideal to use the color legend as a filter.

Pre-Filter vs Post-Filter (the jedi filter)

Filtering a bump chart is not so straightforward. If we use "State" as a quick filter, the entire ranking will be affected because the data set is filtered before ranking. The "State" filter is a pre-filter acting on the data source. Actually, what we need is post-filtering the ranking result. This can be accomplished by creating a magical Jedi filter:

Lookup(min(State),0)

This is a calculated field created from the dimension "State". Using the jedi filter, we are able to select to view or highlight just a few states of interest. We name it "State Filter".

In most cases, we can directly drag the jedi filter to the filter shelf. It can be used as a single-value or a multi-value quick filter. The viz is as follows where we can select to view one or a few of states:

In our case, we will use an alternative filter by parameter, because we want to exclude a blank option due to the fact that some of the causes are major only in a few states.

Thus, we need to create a parameter "Choose a state" that acts as a drop down menu:

And a boolean filter "State filter calc" matches the parameter with "State Filter".

And we will get the following viz that allows us to view one state at a time:

Using circles in Bump Chart

Lines or spaghetti are most used in bump chart. They can be cumbersome sometimes. In the US death cause project, I found that circles or big dots can be a cleaner alternative mark to lines. See the resulting viz as follows:

Hopefully, the technique presented here will expand the usage of bump chart in circumstances where there are a lot of data to be ranked. We can do it by filtering those spaghetti.