Thursday, November 29, 2012

Picking on Clippings

Some time back I saw this beauty and had to snap a picture of if for petty scrutiny later (like this and this).  It is a printed graphic illustrating the relative proportions of 5 major wireless providers' customers, particularly how a potential merger of two smaller providers would compare in that regard.  And it has some problems.

Five numbers represented in a confusing and unnecessarily complex fashion.

At first glance I thought Verizon's magnitude was a small sliver compared to the others, which is unfortunately the exact opposite of the intent.  Verizon is the biggest player in this data set but the chart's first impression makes it appear as a minuscule marginal player.  Nested area charts are at least better when they are unfilled so misunderstandings due to occlusion are reduced.

Nested radial charts that are filled have problems of occlusion and are misleading. Coloring only the stroke sort of helps.

Supposing you are Ok with the problem of occlusion, or at least overcome it, why break this up into two separate graphics?  This is an illustration of a handful of numbers, why pull apart a nested radial chart and turn it into a nested radial chart and a pie chart?

Bringing the bubbles home.

Here's a clumsy reinsertion of the parts into a single graphic where the proportions are still hard to compare, but less so.  It's at least a reduction of complexity with no loss of data.  But still, a nested radial chart with a pie slice?  It turns out, humans are waaay worse at at visually discerning the relative areas of circles than we are at comparing the relative height of columns (on the same plain).

Five numbers shown as more readily-comparable columns in one chart.  Also, I put dumb newsprint texture behind the chart in a halfhearted whim of fairness.  'Fine, I'll isolate the variables,' he growled reluctantly. 

So here it is as a boring old column chart.  It avoids the miscue of Verizon looking like a small portion of the whole, the relative heights of the columns are easier to compare than the relative areas of bubbles, and the hypothetical merger column of Metro PCS and T-Mobile are in a stacked column living in with the rest of the data for easier visual comparison.  Does my labeling look awful?  Sure!  But check out those sweet sweet columns!

Tuesday, November 27, 2012

The Real Losers of Election 2012? The Colorblind

So while this now-entrenched notion of states, counties, people, being either red or blue is (ironically) thematically patriotic, it must be a real drag for the colorblind.  Every election-related map or chart is using these colors, and often the full gradient in-between.  This includes me.  And while I thought about folks with hue discrimination problems, I charged ahead with abandon because the cultural association of red and blue in the political domain was too strong to resist.
And I'd do it again (he said with awkward shame).  But, I'd also make an alternate version again, too.  So here is that alternate version that I thought I'd get to a lot sooner...

Not so hot to me, but hopefully more useful to those with deuteranopia and deuteranomaly colorblindness.  Here, the great American mixing is green rather than purple.

Even still, the dense clouds of 'safe' blue and 'safe' gold mix together you get a  problematic third color: green.  But, ideally this mix color would still look different than either of the Obama or Romney colors.  By the way, the two colors were plucked from the really useful palate of "unambiguous to colorblinds" colors at J*Fly.

Empathy Goggles
Vischeck (in addition to my colleague, Irfan) is a great resource for replicating what an image looks like through the visual systems of folks with different types of colorblindness.  I focused on Deuteranope simulations since that (and the more common Deuteranomaly) is the most prevalent.

Anyways, for those of us who are not Deuteran-seers, here's an imagining of what the original Red/Blue map and the Colorblind corrected map look like through this filter:

The corrected map brings out a more discernible 'mix' color from among the Obama and Romney dots for those with red/green colorblindness.

A visualization that relies pretty heavily on a the trichromatic (the ability to see fully in three primary colors) capabilities of the visual system will find some loss of nuance when adjusted to accommodate a dichromatic (two primary colors) visual system.  But in the images above, it's clear that picking a couple colorblind-unambiguous colors generates a much more visually distinct "overlap" color from either of the inputs.

P.S. Speaking of 'di' and 'tri' chromats, Radiolab blew my mind recently in their podcast on colors.  Apparently it is theoretically possible to see as a tetrachromat.  They are out there somewhere, bathing in hues I can't dream in -and I'm malevolently jealous of them.


Friday, November 9, 2012

Election 2012

As more specific vote counts are coming in, a picture of voting distribution, volume, and party proportion can be formed.  Here's a Goldsberry Map of what Election day 2012 looked like...

More purple than not. Posterize.

Here's a navigable version to peer in on the nitty gritty.

Red/green colorblind?  This version might make the variation more discernible.

This is a mixed density visualization of the volume and proportion of votes in the 2012 US Presidential election. Each point represents 100 votes. This method avoids the geo-social visual bias of large geographic areas having small populations overwhelming the overall picture. In this way both the relative volume and geographic distribution are apparent, as well as the partisan proportions throughout. Areas of mixed voting appear as a blended purple cloud, while areas more homogeneously represented appear more red or more blue.
The data comes from a running update of the latest vote returns per county aggregated and made available by the fine folks at @GuardianData in the case of the original image and @Politico in the case of the updated version.  Special thanks to @BrianTimoney for his help.

For print options regarding this map, visit here.