GO FIGURE Different ways of seeing stats What are we to make of all those stories that warn of lifestyle dangers and slap a giant "%" sign in the headline? Michael Blastland introduces the Risk-o-meter to his regular column.
GO FIGURE Different ways of seeing stats What are we to make of all those stories that warn of lifestyle dangers and slap a giant "%" sign in the headline? Michael Blastland introduces the Risk-o-meter to his regular column.
In response to your e-mails after the last column, and the everyday fog of statistics about risks, I've produced what I hope is a better way to see the numbers.
In response to your e-mails after the last column, and the everyday fog of statistics about risks, I've produced what I hope is a better way to see the numbers.
From bacon to booze, risks often make headlines: "CANCER UP X PERCENT IF YOU DO Y" - you know what I'm talking about. So I've devised a simple but different way of seeing stories, with a click-by-click Risk-o-meter. Click through the examples below to see why those percentages easily mislead - and why it pays to ditch percentages and talk instead about the numbers of real people.
From bacon to booze, risks often make headlines: "CANCER UP X PERCENT IF YOU DO Y" - you know what I'm talking about. So I've devised a simple but different way of seeing stories, with a click-by-click Risk-o-meter. Click through the examples below to see why those percentages easily mislead - and why it pays to ditch percentages and talk instead about the numbers of real people.
A BETTER WAY TO SEE THE RISKS
{current} of {total}NEXTBACK
A BETTER WAY TO SEE THE RISKS
{current} of {total}NEXTBACK
Two caveats. First, links between a food and a health hazard (or benefit) do not always indicate a direct cause. Second, some studies have produced different numbers to the ones here. The benefit of moderate alcohol, for example, is at the conservative end of estimates. Some studies suggest it can be far more beneficial.
Two caveats. First, links between a food and a health hazard (or benefit) do not always indicate a direct cause. Second, some studies have produced different numbers to the ones here. The benefit of moderate alcohol, for example, is at the conservative end of estimates. Some studies suggest it can be far more beneficial.
We'll bring out the Risk-o-meter if stories like these have drifted into abstract meaninglessness and we'll convert them into something more human.
We'll bring out the Risk-o-meter if stories like these have drifted into abstract meaninglessness and we'll convert them into something more human.
To be fair, governments and epidemiologists (people who study patterns of illness and their causes) do need to know the percentages - but don't need the media to tell them. The rest of us, who might rely on the news for guidance, deserve something more relevant.
To be fair, governments and epidemiologists (people who study patterns of illness and their causes) do need to know the percentages - but don't need the media to tell them. The rest of us, who might rely on the news for guidance, deserve something more relevant.
• Go Figure hopes from time to time to bring you arresting ways of seeing data. How the numbers look often changes how we understand them. New ideas for data visualisation - like animation in graphs - are causing a buzz among people who use lots of stats, among them the news media. Not everything we offer will be cutting edge, but we hope it will stimulate.
• Go Figure hopes from time to time to bring you arresting ways of seeing data. How the numbers look often changes how we understand them. New ideas for data visualisation - like animation in graphs - are causing a buzz among people who use lots of stats, among them the news media. Not everything we offer will be cutting edge, but we hope it will stimulate.
Here's an idea (see image below) called parallel sets, designed to show data with categories that divide into other categories. This one shows passengers on the Titanic, male and female, by class of ticket, and by survival, and is designed by Robert Kosara of the University of North Carolina. Take a moment to explore it and let us know what you think.
Here's an idea (see image below) called parallel sets, designed to show data with categories that divide into other categories. This one shows passengers on the Titanic, male and female, by class of ticket, and by survival, and is designed by Robert Kosara of the University of North Carolina. Take a moment to explore it and let us know what you think.
Add your comments on this story, using the form below.
Add your comments on this story, using the form below.
As a visual learner I thought this was a very clear and easy way to read and understand the data. Brilliant!Sue Baker, Leek, GB
This is the first time I've seen this way of representing sets, and it's nicely intuitive. However, the vertical axis introduces a mental bias - the impression that travel class was the most influential factor in outcome, which may or may not be true, given the information displayed. I also would suggest it would make more sense to keep the most coincident sets lined up vertically ie: move the male dominated crew category to the far right and move the survived category dominated by females to the left. Lisa K, Durham NC
I applaud you innovative risk visualisation approach. I have seen friends and family drop needed medications or change diets because of a news story that discusses substantial percent risk changes that only if you investigate turn out to be changes in low overall risks. Your idea of displaying overall risk is right on and might help people key in on changing behaviour related to their highest overall risk factors. Bravo.John Shanklin, Shoreham, New York
It is visually a beautiful graph but lacks an immediate content message. It may work for technical presentations, but not for the average reader. A graph should help CLEARLY and QUICKLY convey a message better (or more efficiently) than text would. This is asking too much of the reader because it takes too much time to understand the message. It may work for a technical paper where you are interested in the analysis, but not necessarily in a mass-media publication where the reader is largely interested in gaining awareness.Ruben Moreno, Oakton, VA USA
Florence Nightingale's original pie charts were much easier to read than this.Caroline Keef, W Yorks
Really like the parallel sets graph, took me about five seconds to work it out, then I realised it's a really good way to present complex data sets. When will my spreadsheet software be able to produce one of these for me?Simon, Bristol, UK
I have absolutely no idea how to read that graph. My guess is as follows - it looks at though possibly the highest percentage of survivors within a group/class is if you were travelling in first class (looks as though roughly a third, or just under a half survived)?Paul Spargo, Cardiff, Wales
I took a little time to understand the image. It is so unlike any data presentation I personally have seen before that my first reaction was to be puzzled. After a minute or two to read the headings, it wasn't at all hard to understand and I thoroughly enjoyed following the various strands. I found it gave me a lot more information in one place that I normally see in reports and gave me the opportunity to start linking variables together and getting a better overall picture of the data. J Howe, Bolton UK
I like the way you can look at the dynamic from different perspectives like "what type of people survived" or "did all of first class survive and everyone else perish". A surprise for me is the sheer numbers of crew compared to passengers and the chances of survival of a male first class passenger was about 1 in 4. Remarkable way of showing the data and engages debate. I think I'll take a copy.Dave Tyrer, Leeds
Visualisation is the key to understanding complex (and somewhat obscure) dataset, especially in statistics. This graph demonstrates that if you really want to commute by Titanic you better be female and pay an hefty fee for your ticket.Armando Forlani, London
It took a moment to understand the Titanic chart but once it "clicked" it made good sense. Its hard to put that much data in one chart and it was an intelligent and clever way to do it. I also liked the way the risk data was presented in the story. Risk is sometimes a difficult subject to explain as there is often more than one reason for the outcome (similar to the Titanic data). Joel Baumbaugh, San Diego, California
I remember a geography professor in a lecture deliberately misusing statistics: "70% of Finland is forest & 35% of Finland is water. This does not leave much room for the Finns."Duncan Peet, Edinburgh
Great way to show stats. But how about actual figures? It tells me that third class male passengers were way more than female passengers and most of them did not survive. But there is no clue as to how many male passengers actually survived. Don't parallel sets display actual figures and numbers? Sourav Basu, Salt Lake City, Utah
Very interesting way of showing multi-faceted data. Takes a little time to figure it out but worth the effort. It would be improved if you could mouse over or select in some way so that one group became highlighted. Or perhaps show the numerical data in a similar fashion? Geoff, Caledon, Canada
I studied statistics as part of my degree in the late 70s, qualified as a chartered accountant and am now a finance director. All this has taught me to ask the right questions when statistics are quoted at us. Too few people question, or are in a position to question the stats. What questions were asked? What was the population? What was the sample? Is the pronounced interpretation of the result the only one that can be placed upon it? I have now got my children questioning the information they are fed: and this is where education really needs to get children thinking, so that they ask the right questions and not just repeat the information given as if it is "the truth". As for this chart, it took some understanding but I liked it and it is a useful way of showing certain types of multi-layered data in a two-day chart. Interesting.Richard Wyatt, Huntingdon, Cambridgeshire
Interesting article proving that percentages really mean nothing unless put it to context. Every week a new statistic (normally as a percentage) comes out telling us we are all going to die sooner, but as shown above the effect is normally very small on a personal level. On another note the parallel sets diagram is an amazingly concise way of showing the data, I may start using it myself.Aly, Scotland
