Archive: Charts and Graphs
February 20, 2010, 10:42 am
Visualizing More Affordable Care
by Henry Woodbury
The February 2010 issue of Obstetrics & Gynecology features work by Dynamic Diagrams for an article titled Alternatives to a Routine Follow-Up Visit for Early Medical Abortion. The article describes a protocol for assessing a woman’s health after an abortion without routine use of ultrasonography. To quote from the abstract:
We constructed five model algorithms for evaluating women’s postabortion status, each using a different assortment of data. Four of the algorithms (algorithms 1–4) rely on data collected by the woman and on the results of the low-sensitivity pregnancy test. Algorithm 5 relies on the woman’s assessment, the results of the pregnancy test, and follow-up physician assessment (sometimes including bimanual or speculum examination).
A sponsor of the study, Gynuity Health Products, asked Dynamic Diagrams to visualize the data. Our explanation shows the results for the current standard of care and five algorithms tested by the researchers. For each approach we show the total number of cases, the number of women returning to a clinic for a follow-up visit, and the number of women receiving a follow-up ultrasound. In contrasting colors we show specific additional treatment cases in two columns; those identified by the protocol on the left vs. those not necessarily identified by the protocol on the right. In large type we provided the percentage of the number of follow-up ultrasounds to the total number of cases. This combination of rich data points and a key percentage makes it easy to compare the effectiveness of each algorithm. A sample of this visual language (without labels) is shown below:
While we cannot reprint the full text of article, we can provide the visual explanation used as Figure 2: Algorithms identifying women who received additional care after medical abortion (PDF, 409K).
February 2, 2010, 9:43 am
Rendered in Neat Circles
by Henry Woodbury
Popular Science links to another interesting information graphic on space exploration. This one, designed by Michael Paukner, illustrates the number of human-created objects orbiting Earth — and assigns responsibility:
You can view larger versions on Paukner’s Flickr page.
The title of my post comes from the Popular Science URL: see-space-debris-cloud-surrounding-earth-rendered-neat-circles. Ironically, this summarizes the problem with the visualization. Despite the attractiveness of the graphic, the neat circles show linear values by area, making precise comparisons completely impossible.
The donut shapes created by the overlapping circles also confuse comparison. Take a quick look at the darkest circles– that for space debris — around the United States and Russia. The United States is bigger, but by what order of magnitude? We see a lot more black — a thicker torus– but the actual ratio is just 1.2 to 1.
December 22, 2009, 11:18 am
Mashing Up Suggestions
by Henry Woodbury
In The New York Times, IBM scientists Fernanda Viégas and Martin Wattenberg have some fun with search engine auto-suggestions. Type in even a single word and you receive “a list of suggested, presumably popular completions.” (In courtroom dramas, this is called leading the witness.)
The fun is seeing how different investigations overlap. Here’s one example:
December 14, 2009, 10:59 am
The Periodic Box of Chocolates
by Henry Woodbury
People seem to forget that the periodic table is a table because it reads in two dimensions. Read it left to right and atomic weight increases. Read it top to bottom and you find elements with similar properties — for example, the alkali metals in group 1 or the noble gases in group 18. The gaps in periods 1, 2, and 3 represent physical realities about the electron configuration of those lighter elements (see this Periodic Table by Chemicool).
Most attempts to fit other data sets to the periodic table result in strange confections.
This Periodic Table of Visualization Methods is a prime example. A simple categorized list is puddled into the matrix of Dmitri Mendeleev’s table and shoved around to fit. There are exactly six “compound visualizations.” How serendipitous. The really interesting data — the examples of the methods — are hidden under reductive two-letter acronyms, making comparison impossible even when you do find something interesting.
If the categories are meaningful and not just quantified to fit the table, the next step is to abandon the presentation method that doesn’t work and come up with one that does.
November 6, 2009, 4:04 pm
Making Your Data Intuitive
by Lisa Agustin
How can we make data intuitive–that is, so it “hits home”? We’ve posted previously on the technique of making large numbers meaningful by using a clever or shocking image. But this method has its limitations. A big number explained with a visual analogy may get people to say “Hey, you’re right, that IS a big number.” But in order to get the audience to act (rather than just react), it takes extra effort to translate that statistic into something they can relate to on a personal level.
Consider the funding coming through the U.S. government’s Recovery Act: $787 billion. Sure, that sounds like a lot of money. But is it too much? Too little? It depends. Authors Dan Heath and Chip Heath explain it this way:
How can you relate to this monstrous figure in the daily-life zone? Well, there are roughly 112 million households in the United States, with a median household income of about $50,000. So an $800 billion stimulus works out to be the rough equivalent of seven weeks’ income for an American household. Is that worth it? By way of comparison, we already work three or four months a year just to pay our federal, state, and local taxes. So maybe this seems like a no-brainer to you: seven weeks’ worth of work to stave off a potential depression. Or maybe you’re appalled. Regardless, we can finally have a real argument, because we have a better idea of what we’re arguing about.
Well said.
October 30, 2009, 3:39 pm
Hey Jude, Don’t Get Confused
by Henry Woodbury
Created by love all this (hat tip to Sippican Cottage).
October 16, 2009, 10:10 am
Infographics for Web Workers
by Lisa Agustin
Web Design Ledger offers a collection of infographics of special interest to web workers, including process flows, data driven visualizations, and musings (like xkcd.com’s Map of Online Communities, above). Enjoy.
October 15, 2009, 9:22 am
What Are the Odds?
by Lisa Agustin
Just out this week, the Book of Odds claims to be “the world’s first reference on daily life.” Normally, I’m not too interested in finding out my odds of surviving a plane crash or ever having eaten pizza for breakfast, but with its broad collection of statistics, articles (“Behind the Numbers: the Sharks and the Vending Machines”), and a personalized feature for creating your own book of odds, the site makes for a fun diversion. Browse statistics by area of interest (Accidents & Death, Daily Life & Activities, Health & Illness, and Relationships & Society), or use the Visual Browse tool to view odds on a keyword of your choice. But don’t let the title fool you: while the site is about numbers, it doesn’t offer gambling or predict the future (too bad).
September 25, 2009, 3:32 pm
Data in the Round
by Henry Woodbury
An interesting, but flawed chart at O&G Next Generation shows how much oil the United States imports from other countries:
There are several big problems with this chart. First, U.S. oil imports per day by country is linear data. When one-dimensional values are presented as two-dimensional areas, proportional differences between values are rarely perceived correctly. This problem is compounded by the placement of the data blobs on the global map. It is good to attach each blob to a country, but not good to scatter them both vertically and horizontally. With a little design attention the values could be presented as bars and aligned along a single x-axis in the tropics.
Another problem is that several important data points aren’t shown. Most importantly we need a figure for the United State’s domestic production. This is vital for context. Upon investigation, we find that the bar chart on the bottom left is either not accurate or not tracking the same petroleum product as the map. If you subtract Total Imports from U.S. Consumption for 2008 you get a ballpark figure of around 6,000 thousand barrels per day. This is far off the mark. The real number for 2008 is 4,921 thousand barrels per day, a little bit less than total U.S. crude produced since a small amount of U.S. crude is exported. In June 2009, domestically produced minus exported crude is 5,126 thousand barrels per day.
Another missing figure is the total of oil imports from all countries after the top 10. Once we can look up the June 2009 total for all countries — 9,172 thousand barrels per day — we can easily calculate the sum of all countries after the top 10. The long tail total turns out to be 1,613 thousand barrels per day which is greater than all but Canada. The 9,172 total and various subtotals also allow us to validate the 82% percentage on the far right and update the 2007 ratio of 60% to the actual June 2009 ratio of 64%.
If we add circles to show the oil consumed by the United States from its own production and the “Rest of World” total identified above, the chart looks something like this:
It is even more difficult to read. But that’s not a problem with the data. The data needs to be shown. The problem is with the presentation. The chart still shows linear values with areas, it still doesn’t show totals, it still uses an out-of-date figure from 2007 on the far right, it still has questionable, out-of-date data on the bottom left, and it still has a jumble of factoids on the bottom right that don’t relate the data above. Alas, I am out of time.
September 8, 2009, 1:02 pm
The Max Baucus Health Care Lobbyist Complex
by Lisa Agustin
The current health care reform debate has presented plenty of opportunities for visual thinkers (and aspiring ones) to clarify the issues and explain possible solutions. My current favorites have been Dan Roam’s “back of the napkin” series on fixing health care and the flow chart prepared by the office of Congressman John Boehner (R-OH) showing the Democrats’ health care proposal. (Should we assume that the awfulness is on purpose?).
But subtler visualizations grab my attention more for what they imply. Consider “The Max Baucus Health Care Lobbyist Complex,” which was developed by the Sunlight Foundation, a group whose goal is to “use the power of the Internet to shine a light on the interplay of money, lobbying, influence and government in Washington in ways never before possible.” The Max Baucus visualization is named for Sen. Max Baucus (D-MT), who heads the Senate Finance Committee, which has been singled out by advocates and news organizations as the toughest obstacle for the President’s health care priorities. The visualization shows the connections from Baucus to five of his staffers-turned-lobbyists to their health care sector clients, which, in some cases, overlap. Most of the organizations are directly involved in the health care or insurance industries.
According to the Foundation:
In his many years on the committee, Baucus has amassed a wealth of connections to the health care and insurance industries, often through his ties to former staffers turned lobbyists. These connections expose how close the many organizations seeking influence on health care reform are to one of the most powerful players in Washington.
Data for the visualization was provided by OpenSecrets.org.
September 4, 2009, 1:12 pm
The Times Goes Google on Us
by Henry Woodbury
I just discovered the New York Times Developer Network.
This resource provides data from The Times to third party developers through content-related APIs:
Our APIs (application programming interfaces) allow you to programmatically access New York Times data for use in your own applications. Our goal is to facilitate a wide range of uses, from custom link lists to complex visualizations. Why just read the news when you can hack it?
Most or all of the APIs respond to a query by returning data in XML or JSON format. Some developers have built custom search engines and topic-specific mashups around this functionality. Others are more interested in the sheer excess of the data — and how it can be visualized.
Artist Jer Thorp is one of the latter. Thorp accesses the Times Article Search API to create visualizations that compare the frequency of key words over time. The image below, for example, compares ’sex’ and ’scandal’ from 1981 – 2008:
When you zoom in, the visualization reveals branching segments called “org facets”. Thorp writes:
[These are] organizations which were associated with the stories that were found in the keyword search. This is one of the nicest things about the NYTimes API – you can ask for and process all kinds of interesting information past the standard “how many articles?” queries.
September 2, 2009, 2:20 pm
What’s Wrong with this Chart?
by Henry Woodbury
The chart, of Federal Spending FY 2009 YTD, is from USAspending.gov, a web site mandated by law to provide the public free, searchable information about U.S. Federal expenditures.
Seth Grimes at Intelligent Enterprise figures out the problem and its cause:
USAspending.gov produces its charts dynamically using the Google Chart API…[but] passes values to Google that are out of range. Google truncates them, just as [its] documentation explains.
Here is Grimes’ corrected chart:
Unfortunately, data misrepresentation isn’t the only problem he finds.
August 24, 2009, 8:52 pm
The Pint Slide Rule
by Henry Woodbury
This is a not a post about a beer gauge. It is a post about cognitive bias. To quote from the item itself: “[Jean] Piaget studied the tendency to focus attention on only one characteristic. In our case: beer height not volume!”
The gauge is the invention of engineer and physicist Chris Holloway. The Wall Street Journal Numbers Guy, Carl Bialik, explains:
Holloway has noticed that the typical pour in a pint glass is less than a pint. And since the widest part of the glass is at the top — nearly twice as wide as the bottom — leaving just the top half-inch of the glass unfilled costs the customer nearly 15% of the pint he’s paying for. So what may look trivial to bartenders and to drinkers, thanks to our tendency to focus on height rather than width when taking the measure of liquids, is a serious tavern injustice.
Readers of Edward R. Tufte will remember that one grave mistake in visualizing data is to show one-dimensional data with two-dimensional graphics:
There are considerable ambiguities in how people perceive a two-dimensional surface and then convert that perception into a one-dimensional number. Changes in physical area on the surface of a graphic do not reliably produce appropriately proportional changes in in perceived areas. The problem is all the worse when the areas are tricked up into three dimensions. (The Visual Display of Quantitative Information, 2001, p. 71.)
Holloway had the reverse challenge — the problem people have perceiving differences in volume. The gauge turns three-dimensional data into a one-dimensional series. It is a portable liquid measure.
I will add that despite the elegance of the Holloway’s concept, the gauge as is could use some design improvement. Holloway has elected to emphasis even numbers of ounces. I’m not sure this helps readability. There’s also no reason to have the 6 oz. label offset from the 6 oz. line. Just make the card a quarter inch longer or so. Nor is there reason for the key or the “Glass edge” and “Beer surface” labels. Instead, replace all of this extra text with a 16 oz. line aligned to the top of the glass that incorporates some minimal description. For example:
16 oz beer / 0% missing
15 oz / 7%
14 oz / 13%
etc.
August 19, 2009, 9:50 am
You Are…
by Matt DeMeis
MIT Phd student Aaron Zinman has created an interesting data driven visualization experiment called “Personas”. Simply enter your name and a Flash app scours the web for bits and pieces of information about you. As it does so, its progress is displayed in visual form (albeit at warp speed, so it’s more for “ooh ahh” factor than usefulness). You are then characterized as a colored strip of categories ranging from books, sports, management and aggression to education, legal and illegal (activities?). It’s an interesting experiment. I think it would be great to have a bit more control over the categories and info about the user. Just to help weed out the cruft. As is, it’s probably pretty inaccurate for someone with the name Bob Smith or Michael Jackson. I am curious what our resident Flash maven Piotr would add to it. Go try it out over at the MIT site.
June 30, 2009, 1:00 pm
How Tall is the Green Monster?
by Henry Woodbury
Flip Flop Fly Ball is Craig Robinson’s collection of “baseball infographics”:
Essentially, this site is what I’d have been doing when I was 12 years old had the Internet and Photoshop been available to me in the eighties.
What stands out for me from this collection is Robinson’s ability to ask good questions — intriguing or amusing or both.
In some of the work, the question is more the point than the answer. What if baseball players literally stole bases? For more complex questions Robinson often produces just a well-drawn pie or bar chart. But occasionally, Robinson combines question, data, and visual idea into a smart visual explanation that goes beyond that.
For example, the left field wall in Fenway Park is 37 feet and two inches tall. And how tall is that?
April 22, 2009, 8:27 am
Broken on Purpose
by Henry Woodbury
Seth Godin at Gel 2006 explains how This is broken. What is broken? Almost everything.
Including Napoleon’s March to Moscow.
Starting at 17:53, Godin buries Edward Tufte in order to praise him. Note that Godin doesn’t really bother with the graph itself, but rather Tufte’s promotion of it as “the best graph ever made.” Godin responds:
I think he’s completely out of his gourd and totally wrong!
If you need to spend 15 minutes studying a graph you might as well read the text underneath. Godin then backs off. Tufte’s promotion of Napoleon’s March, he says, is an example of something “broken on purpose”:
For the kind of person you want to reach — they want to read a complicated difficult to understand graph and get the satisfaction of figuring it out, because then they get it…. Sometimes the best thing to do is break it for the people you don’t care about and just make it work for the people you do.
Agree?
Watch the rest of the talk as well. It’s a very funny, pointed critique of bad information and product design.
March 11, 2009, 8:18 am
Visualization a la Dilbert
by Lisa Agustin
Proof that even Dilbert understands the power of visualization.
February 3, 2009, 3:55 pm
How We Read Graphs
by Lisa Agustin
Cognitive Daily offers up some recent research into how we read graphs, including some results gleaned from using an eye-tracking device, a tool more commonly used in evaluating web sites.
December 29, 2008, 9:17 pm
Visualization Takes Its Toll
by Lisa Agustin
Sometimes too much visualization is not a good thing. Just ask xkcd.com.
December 3, 2008, 9:45 am
I Heart Coffee
by Henry Woodbury
Christoph Niemann brews up a brilliant illustrated essay on one man’s history with coffee. Don’t miss the chart on coffee-bias-over-time about halfway through (oh sure, it could be improved, Tuftelike, but that’s not the point).
November 20, 2008, 11:34 am
Worldwide recession? Time to make a zombie movie.
by Lisa Agustin
From science fiction site io9: a chart that correlates periods of war and social unrest with the production of zombie movies.
November 3, 2008, 3:48 pm
Microsoft Chart Advisor — Consider the Source
by Mac McBurney
The prototype Chart Advisor for Excel 2007 from Office Labs sounds like a step in the right direction:
This add-in uses an advanced rules engine to scan your data and, based on predefined rules, displays charts according to score. Top scoring charts are available for you to preview, tweak, and insert into your Excel worksheet.
An early post by Program Manager Scott Ruble describes the Excel team’s motivations, which at first glance seem admirable. On second thought, Ruble’s understated description of the group’s noble “intent” and responsiveness to strong feedback reminded me not to get my hopes up. (Emphasis and sarcastic comments added by me):
When Office 2007 was released [and not before then?], one of the strong pieces of feedback was Excel needs to do a better job guiding users in the proper selection of charts to effectively communicate their data. Though it wasn’t our intent [I feel so much better now], some of the new [and the old] formatting options [and defaults] such as glow and legacy 3D charts can [only] be used inappropriately, which obscure[sic] the meaning of a chart. Some people [silly, silly people] felt that these features contributed to creating more “chart junk.” In an effort to improve this situation, we have created a prototype called the Chart Advisor.
Mr. Ruble is being too modest. The new features and default settings — like the old features and default settings — guarantee more chart junk. This team wasn’t born on the day Office 2007 was released — quite the opposite. Saying that inappropriate use and obscuring the meaning of a chart was not the team’s intent seems, frankly, laughable.
I expect an upgrade from Microsoft to include new features — new things that users could do. Giving good advice about what a user should do is more difficult and risky, and it would ultimately be much more valuable. This is ambitious, and let’s hope it signals a greater focus on improving the real-world capabilities of Excel users, not just increasing the capabilities of the Excel software.
So far, Chart Advisor is in no danger of becoming an artificial Edward Tufte inside Excel. The add-in still serves a side order of chartjunk with your data.
Tim Mays reported that Chart Advisor ignored a whole column of source data and then (not surprisingly) recommended the wrong chart type. At first, Excel guru Jon Peltier didn’t even get that far.
Hey, that “advanced rules engine” is just a prototype. (More on the rules engine). If the wizards at Microsoft succeed in upgrading Excel’s brain, here’s hoping they have the courage to give it a heart and good taste as well.