I did this chart in a few hours as a response to a question on Quora:

"How can I see (or make) a chart that shows household income in the United States (including capital gains) by decile (or even better, by percentile)? So the x axis is people (deciles or whatever) and the y axis is dollars. Since there's such a big spread between the lowest and highest incomes, maybe the y axis should be logarithmic?"

This is a familiar problem in visualization. How do you handle charts when some of the values are so big they swamp the rest? Using logarithms is often not a satisfying solution; many non-technical users don't understand it and as a result the effect may be to conceal something which should instead be emphasized.

One alternative is to embrace the outliers and let them simply smash through the top of the chart. You can then replace the missing data by showing a thumbnail inset. This is what I attempted to do in this chart.

I got a CSV file of income broken down by tenth of percentiles from the University of Minnesota’s IPUMS-CPS project via the website DQYDJ.net. Income is defined by the 2014 whole year INCTOT variable, for people reporting 1+ hours usually worked in 2014. For more information about how INCTOT is defined, see https://cps.ipums.org/cps-action/variables/INCTOT#comparability_section

Creating a chart like this in NodeBox is fairly straightforward. I had to add a few extra nodes to show the first two negative income values below the X axis. I then colored everything above 95% blue and created a thumbnail by simply scaling my original chart by 10%. Because the highest income bar was so faint in the thumbnail, I added a slightly thicker line for emphasis.

I am attaching the NodeBox network, the CSV file, a screenshot, and the PDF output. Comments appreciated!

John