2
$\begingroup$

Sankey diagram is good way to visualize multidimensional contributions (i.e. data having the same sign and contributing to a total).

Is there a similar way to visualize multidimensional change data? E.g. I want to analyse the change in the federal budget of USA (2017 vs. 2016), more specifically the revenues. How can one visualize the change and the interactions by source, state and industry dimensions in one plot?

As an example the plot should be able to show "the federal revenue from NY has decreased by USD 20mn, the change is coming from Financial Industries decreasing USD 40mn, compensated by a USD 20mn increase in Construction". One could have more than 2 dimensions.

$\endgroup$
  • $\begingroup$ Percent change? absolute change? How finely divided the budget? Do you want to show changes at several levels? (as is total defense budget only, or by navy, usaf, ... in addition, possibly multiple levels?) $\endgroup$ – kjetil b halvorsen Jun 26 '18 at 15:15
  • $\begingroup$ @kjetilbhalvorsen The total for time being. I want to be able see at multiple levels. E.g. i want to visualize "the federal revenue from NY has decreased by USD 20mn, the change is coming from Financial Industries decreasing USD 40mn, compensated by USD 20mn increase in Construction". $\endgroup$ – teucer Jun 26 '18 at 15:23
  • 1
    $\begingroup$ Can you please add this information to the original post? $\endgroup$ – kjetil b halvorsen Jun 26 '18 at 16:35
0
$\begingroup$

One option is simply a table. You could format the cells to show the degree of increase or decrease, and you could use the margins to show state and industry totals. The advantage here is that you can fit a lot of detail into a single plot, but the eye is naturally drawn to the biggest contributors (either positive or negative). In this toy example, the NY Financial Services are doing quite bad, and VA as a whole is performing poorly.

Revenue Table


A second option is a waterfall chart, which I like to use to show the cummulative impacts that produce a change. Listing all states and industries separately will make the plot quite busy, so it may make more sense to have one chart per state.

Revenue Plot

$\endgroup$
  • $\begingroup$ these are suitable if you look dimension by dimension. Is there something to also see the interactions? $\endgroup$ – teucer Jun 28 '18 at 14:34
  • $\begingroup$ In your example with two variables, the interactions are shown in the cells in the table example (as opposed to the margins). Perhaps if you provide a third dimension with example data, I can try to fit it into the answer. $\endgroup$ – Underminer Jun 29 '18 at 16:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.