Hard work pays off

At the NY Tech Meetup, Andrei Scheinkman showed off some work his team at Huffington Post did relating to gun violence in America.

 

Interactive version is here. The animation shows day by day, where the victims of gun violence were located. The table below contains the details of each victim, and links to the news story covering the event.

***

What is not seen on the chart is even more impressive. Andrei described how they looked around for databases that would provide them the raw materials for creating this chart but no timely source exists. This means that a team of 15 (if I heard correctly) spent a month or so manually collecting all the data on a spreadsheet.

It's also the reason why they cannot continue the map indefinitely, as people have other things to do.

Andrei also contrasted this visualization with a text article that describes the state of gun violence in words. You guessed it, the visual presentation is hands-down more compelling.

Comments

[Frank Marshal Davies]

Of course this visualization is more compelling: the choosen "bloody" color, size of bubbles that seem to capture the whole country in a litteral bloodbath all cause strong emotions.

This chart is good at causing emotions, but it is a poor way to cut through all the cases to get a better understanding of the state of affairs.

[Mike Liveright]

Of course the other problem is that there is no indication of the Gun Deaths/person.

We know that there are more people in New York than Montana.

Perhaps a Re-coding that showed each death as a Circle of diameter, or Brightness inversely sized or light based on population density would show the effective death risks in various areas of the country?

[derek]

I don't know if they did it, or if my browser didn't pick it up, but sound would also have helped. I'm not suggesting anything tacky like gunshots, just a click for every event.

[r.e. spondent]

I hate bubble / map charts. Are the gun related deaths in Miami really spilling onto the Atlantic Ocean? Are gun deaths in NOLA really spilling into Lucedale, AL? Also, how reliable is this data? How does it compare to UCR statistics or NCVS? Lastly, every bubble or heat density map that is not presented in rates is just an indicator of pop density.

[jlbriggs]

I think people concerned about the population density are vastly missing the point.

The map is an illustration.

What it is illustrating is the fact that in the 98 days following the Sandy Hook shooting, 2,244 people were killed by guns.

How densely populated their town or state was is immaterial. Clearly, where there are more people, more things (of any nature) happen.

The illustration serves a purpose of underscoring the data, and it serves it well. Regardless of opinions of guns or gun laws.

The idea that the color is overly dramatic is pretty silly as well. We use red in charts very often to highlight things that are bad, that require attention, or that are an area of danger.

Again, regardless of opinions on guns or gun laws, deaths are obviously something we can consider bad, or an area of concern/danger.

As for bubbles on a map in the first place - I don't see how it could be argued to be a poor choice for displaying quantitative information geographically.

It requires willful ignorance of the display medium to think that it means the outer edge of the bubble marks the place of the occurrence.

[ascheink]

Hi Kaiser, thanks for the writeup. Fwiw, here's the video I showed of the animation on our homepage http://youtu.be/sv2YZpZWSCY

As to the comments about population density, it's something we thought a lot about when we put this together. We considered aggregating deaths by state and showing per capita rates. We also looked into matching up incident counts with state-by-state gun laws and with ATF gun shop location data.

In the end, we decided to stay away from aggregate analysis and instead emphasize the individual tragedy of each death (which is why we include links to the underlying news stories) and the staggering number of deaths over time.

[Jeff]

The population density issue is a red herring here, as jlbriggs points out. The point of the map is not really to talk about the distribution (or, for that matter, to support much of any kind of analysis); it's to place the deaths in a context that people can relate to, in order to support the obvious point of view of any article that seeks to put human faces on a few who have died. Of course the color and bubble size support this message, too. It may not get us closer to understanding the state of affairs from a quantitative viewpoint, but as jlbriggs says, it's an illustration.

If you want the numbers to help place Sandy Hook in context, the more compelling element is the bar chart at the bottom. My only suggestion here would be to have included the Sandy Hook deaths on 12/14, encoded somehow to differentiate them from the others on that day. This might serve as a kind of scale, and make the point that we've seen worse days for gun deaths since then, even including them.

[zbicyclist]

The population density isn't really a red herring in my opinion. As it stands, the more solidly red areas clearly "tell me" that areas with strong gun control laws (such as Illinois, which was the last state to allow concealed carry) are more red than gun-loving Wyoming. That isn't what this says.

I like Jeff's idea (his second paragraph) to show that the days with media-worthy gun events don't really stand out that much from days with ordinary gun violence.

[Ken]

As they make the data available, it would be no trouble to construct a map based on deaths per population.

[r.e. spondent]

A bubble chart on top of a map is a two dimensional presentation of data. Using that same logic, if I overlaid bubble of job loss over years, the 2007 & 2008 bubbles would spill over to 2006 and 2009 which makes false implications.

I guess my problem with this is that the scale of the map conflicts with the scale of the bubbles. The Chicago bubble is bigger than the state of Illinois.

[jlbriggs]

@r.e. spondent

I think you are missing the basic concept of a bubble as a display of data.

The only part of the bubble that correlates to the geographic space is the bubble's center point.

The area of the bubble is obviously not intended to show the extent of the geographic distribution. Were that the case, a choropleth map would be used. Just as it is obvious that the overlapping bubbles do not mean that the people represented in those bubbles died stacked on top of each other.

A bubble is - in general - a poor way to display quantitative data because it relies on area which is difficult to compare.

However, when the need (or want) to display quantitative information on a map arises, the standard methods do not work well (bar charts displayed geographically, for example, have no common baseline for comparison. In cases where a bar chart would overlap another, they would both be entirely useless).

In such a case, the less than perfect stand alone bubble becomes one of the better options for encoding the data.

[...........]

Wouldn't a heat map be better than this bubble chart with overlapping bubbles? Either a chloropleth of counties if that data is available, or else some kind of kernel density estimation if only nearest city data is available?

[jlbriggs]

A heat would face the same drawbacks that a bubble does, while possibly being more specific geographically.

And again, I think the main point is being missed here...the exact geography is not the main point...it is an illustration used to emphasize the data - the important part: the number of deaths.

[Kaiser]

Great to see this discussion here. The absolute versus relative value is a decision that crops up everywhere. I often find that looking at both things side by side is the most valuable. However, that would be awkward for a journalistic piece.
As it stands, Andrei's team is emphasizing where the deaths occurred. Of course, the big cities have more deaths.
What some of you are uncomfortable about is the limitation of embedding population data onto maps in which the area is not proportional to population.
One idea is to structure the data more. Come up with some regional clustering like big cities, southern towns, etc. and aggregate the data.
But don't forget my main point - this is a visualization in which the designers made sure they have good data. There are countless charts out there in which the designers just plotted whatever they could get their hands on.

[r.e. spondent]

Well said Kaiser. It's awesome to have a space to disucss design issues with visualizing data.

I would still like to see a map that compares NCVS crime estimates against FBI stats and/or whatever ever source this map used.