Bubbles of death
Feb 21, 2007
Thanks to Dustin J for bringing this stupendous chart to our attention. I have to admit I have trouble understanding it. The red curve appears to be part of a gigantic circle confirming that all life do end on this earth. How it is connected to the rest of the chart I am unable to discern. In addition, the trajectory of the bubbles, the overlaps between bubbles, the separation between bubbles all may or may not carry meaning.
Reference: "What are the odds of dying?", National Safety Council.
The best bit is that 'terrrorist attack' is not on the list. Neither is 'killed in military action'. One would have thought that given the fact that the US is currently spending some 1 trillion dollars on the combination of the two topics, that they would at least get mentioned.
Posted by: Jens | Feb 21, 2007 at 05:32 AM
Seems to be some sort of area chart, where the area denotes the chance of death. I think the trajectory and overlap contain no meaning.. cant really make out. I'm also missing out the point that the chart is making. Like you said, its pretty unreadable.
Posted by: Siddharta | Feb 21, 2007 at 05:34 AM
The spiralling pattern and overlapping bubbles are a great case of misdirection. After reading the chart I thought I smelled a rat, and looking at the data table in the NSC article confirms it. For example, ICD X45, "alcohol poisoning" with an estimated rate of 1/10,048 plotted, while ICD X41, "narcotic overdose" with an estimated rate of 1/406 (comparable to "firearm assault") not plotted. What was the motivation of the plotter in choosing the causes depicted?
I also wonder why only one year's data was tablated, when up to 5 years worth is readily available at the CDC's WISQARS site at http://webapp.cdc.gov/sasweb/ncipc/leadcaus10.html - overall this is pretty, but pretty superficial and biased.
Posted by: TheRandomTexan | Feb 21, 2007 at 06:17 AM
I am not sure if everyone agrees, but I see this as an infograph, not a chart. In this sense, I usually expect an aesthetic experience and not only a rigorous display of data. Basic rules should apply, but I don't mind to exchange that aesthetic experience for a bit more cognitive processing. In work-related visualizations I like pleasant displays but not at expense of efficiency or cognitive processing time.
Posted by: Jorge Camoes | Feb 21, 2007 at 06:53 AM
Jorge's right, this is an infographic, with different aesthetics. The circles spiraling around in decreasing order of size is part of that aesthetic. A technical chart would use columns (a pareto chart) or markers (a dot plot) with values decreasing in a linear patter from left to right or top to bottom. The big red outline shows where 100% is on this scale.
Posted by: Jon Peltier | Feb 21, 2007 at 07:29 AM
Oops! The "narcotic overdoses" are actually ICD X42: "Accidental poisoning by and exposure to narcotics and psychodysleptics [hallucinogens], not elsewhere classified."
Another odd omission: the 762 bicycling fatalities of 2003 rate a spot on the chart, while the 15,658 HIV fatalities (see WISQARS) don't. What's up with that?
Posted by: TheRandomTexan | Feb 21, 2007 at 08:13 AM
The spiral and placement of the bubbles look to be purely asthetic, and I don't particularly mind that.
The overlap though is a problem. It takes me a moment to realize it is't meaningful, when I _expect_ it to be. (And, in a couple of cases, it could be.)
I think a pie chart would have been more readable, and almost as nice to look at.
Posted by: Daniel Staal | Feb 21, 2007 at 12:23 PM
There was a similar spiral graph in Sunday's New York Times Magazine here:
javascript:pop_me_up2('http://www.nytimes.com/imagepages/2007/02/14/magazine/18wwlnessay.graphic.ready.html', '595_1039', 'width=595,height=1039,location=no,scrollbars=yes,toolbars=no,resizable=yes')
If that link doesn't work try the link here and click on the "multimedia" link in the left hand column:
http://www.nytimes.com/2007/02/18/magazine/18wwlnlede.t.html?ref=magazine
Seeing two spirals used in three days makes me wonder about the attraction. One graphical reason might be space on the page. But neither works well as a statistical presentation.
Posted by: Charles Franklin | Feb 21, 2007 at 06:18 PM
And, to me, the spiral is backwards. If 100% is the largest circle (red curve), then shouldn't the next largest circle be closest to it, spiralling in towards the center of the circle, smaller and smaller?
Posted by: Merle | Feb 21, 2007 at 07:46 PM
The red curve is an arc whose (apparent) area is five times that of area of dying of a heart attack (100% to 20%).
I realize that these probabilities are all conditional...if I don't swim the chances I'll drown are less than people who spend lots of days on the lake.
...But the probability of dying by legal execution?
Posted by: IB a Math Teacher | Feb 21, 2007 at 08:29 PM
IB: good observation. It's worse than you think. According to the article: "The odds given below are statistical averages over the whole U.S. population and do not necessarily reflect the chances of death for a particular person from a particular external cause." I take this to mean they are not conditional but rather expected values. In that sense, it's correct for no one.
I also take objection to the inclusion of the 1 in 1 chance of dying line. It's a cute addition but really does not help in interpreting any of the odds. If you tell me the odds of dying from fire or smoke is 1 in 1113, telling me further that I can compare it to 1 in 1 really adds nothing to my understanding.
Posted by: Kaiser | Feb 22, 2007 at 12:45 AM
Forgive my naïveté, but it seems to me the categorical options don't even come close to adding up to 1. So the biggest question I'm left with after viewing this chart is, how do the other half die?
Posted by: Seagrinch | Feb 22, 2007 at 11:11 PM
Seagrinch:
Not sure, but I'm estimating that the probability of a person with cancer dying by stroke after severing body parts in a motor vehicle accident (the darkish, darkish, lightish, green area) is approximately 1/43.
Posted by: IB a Math Teacher | Feb 23, 2007 at 10:05 PM
This may be obvious, but the graphic is not what chances an individual has of dying a certain way, but rather if one looks at a single death with no other context, what the probability is that the cause of death was that specific event type.
In other words, it's descriptive only for the total US population. Once you get more specific, I'm sure the statistics can veer off wildly in many directions.
For example I would think (hope) that the chance most people would be killed by a legal execution approaches zero.
I enjoyed the graphic as an aesthetic experience, but I agree the construction is confusing - especially the overlapping of causes.
Posted by: illovich | Feb 28, 2007 at 10:41 PM