Archive for the ‘Statistics’ Category

The Lessons of 2013


Every week I scan the abstracts of about 25 new papers published in peer reviewed journals and by universities. Sometimes I have access to complete papers.

They are all relevant to cycling and I try to stick to the one that have some basis in, or relevance to, science. Considering I read only those written in English, ones that cross my radar and ones that I have any hope of understanding, clearly there’s a lot out there that I miss. Nevertheless, the pickings are rich and diverse.

While I tweet nearly everything I find (@cyclingscience1), here’s a summary of  a little of what I’ve learned this year from those thousands of diligent researchers who continue to add to our understanding of cycling.

I don’t necessarily agree with any of them.

• Yoga stresses the heart and respiratory system less than cycling
• The weaves of skinsuit materials affect your aerodynamics
• Bike reviews criticising comfort are largely untrustworthy
• Regenerative braking for e-bikes is going to blossom
• Cycling in London is either more dangerous or the safety models were wrong
• The Mayor of London is more worried about commerce than road safety
• Mountain bikers suffer the worst injuries in the first third of an endurance race
• French riders in the Tour de France live longer than mere French mortals
• Traffic calming and separate cycle paths make cycling safer in Netherlands
• Medics worldwide believe that bicycle helmets are fantastic
• The health benefits to US society of cycling outweigh the costs
• Caffeine definitely helps if you drink it, but not as a mouth rinse
• Cars don’t pass helmeted cyclists any closer than they pass bare-headed riders
• Steer by wire is on its way for e-bikes
• Support for, and research into, safety in numbers is growing
• Male cyclists have bigger thighs than triathletes
• The secrets of bicycle stability and steering remain enigmatic
• The best time to ride along Oxford Street in London is 10:07 on 25th DecemberOxford St cycling 25 Dec

To stay ahead of the bunch in 2014, buy a copy of Cycling Science and follow the tweets @cyclingscience1

Strava users help sports science – unwittingly


Do you use Strava? If so, you may have contributed unwittingly to pioneering research that could help all cyclists.

Three Italian researchers accessed the data of almost 30,000 Strava-using cyclists. (Put simply, Strava is the social fitness app that tracks your ride and creates a leaderboard for all rides on the same route) The users were anonymous to the researchers so it could be anyone’s data – including your’s.

SocialBlog_In-Situ-1024x592Then the researchers mined this mountain of data. They wanted to see what kind of exercise leads to better performances.

Sports scientists, doctors, physiologists and others have been doing the same kind of research for decades. They’ve come up with a lot of credible theories – but they’ve been based on the results from a few dozen professional or, occasionally, a few hundred experienced participants.

Those old studies look tiny compared to the cohort used by Paolo Cintia, Luca Pappalardo and Dino Pedreschi of the Dept. of Informatics at the University of Pisa, Italy. Thanks to the Strava data, they had an enormous sample size of 29,284 cyclists to study. The vast majority of them would have been amateur (that’s you and me) and their fitness levels would have ranged from the near-elite to the pathetic (me).

In the old days the quantity of data would’ve been too much information to handle easily but, with every second of those riders’ activities in stored in digital form, the researchers were able to drill down relatively quickly.

Fortunately for all sports scientists and coaches, the findings from the huge sample corroborate what’s suspected already from the old, small studies. Exercise on its own doesn’t make you perform better; it’s down to training.

“Athletes that better improve their performance follow precise training patterns usually referred as overcompensation theory, with alternation of stress peaks and rest periods,” say Cintia, Pappalardo and Pedreschi.

“To the best of our knowledge, our study is the first corroboration on large scale of this theory, mainly confirming that “engine matters”, but tuning is fundamental,” they say.

Sadly the potential for data from Strava and other social fitness platforms to help scientists get new insights is now restricted. Paul Mach, an engineer at Strava and creator of Raceshape, implied by tweet that the researchers had only acquired the data because they had “hammered the v1/v2 API before it got shutdown. We blocked an Italian univ[ersity] IP a while back. Probably them.”

Screen shot 2013-11-20 at 15.55.01

This approach by Strava gives the company more control over who can access the anonymised data of its users. “Data acquiring has been a fundamental part of our work,” the Italian researchers say, “We did it through Strava’s API (version 2.0)*. Unfortunately, Strava changed his [sic] API policies in June 2013, so it is not possible to download data anymore.

“At that time, we asked Strava and they were still developing the new version of API. Now it seems that such new version is finally available but you need to request the access to Strava developers.”

Neveindexrtheless, expect more results soon from the Strava data that Cintia, Pappalardo and Pedreschi harvested before the tap was turned off. “Currently we are investigating other fascinating aspects emerging from Strava data, we hope to get new results for helping cyclists in their training life. Specially because we are cyclists, too,” they say.

And if Strava relents and let’s scientists get fresh data, it’s likely to be just one source of information that could benefit all of us, including the most pathetic riders (me) in unexpected ways. “We are sure that the increasing diffusion of training devices (powermeters, heart rate monitor etc) and social fitness applications will give us the possibility of a deep and new study of training science,” say Cintia, Pappalardo and Pedreschi.

Their research paper is available here and they were scheduled to present it in December 2013 at a workshop of the IEEE International Conference on Data Mining in Dallas, Texas


*Update at 16:56 GMT 20/11/13: From Paolo Cintia – “The data acquisiton was done after a request to Strava developer, explicitly highlighting the scientific and anonymous use. Furtherrmore the access to the API was public.”

Update at 09:00 GMT 5/3/14: From the researchers: “We are currently working on some improvements and extension of the cyclists’ study. Our purpose is to create a model able to detect if a person trains in the right way or not. In the meanwhile, we opened a blog where we tell in a divulgative way our scientific works. You can find a post about the cyclists’ study here

More contenders for 2012 Cycling Science awards


Here’s a couple of new nominations for the 2012 Cycling Science award. If you would like to suggest others, please use the reply form at the bottom of this page.

Out for the count

City planners need the right information to make the best decisions for encouraging cycling. Unfortunately the best information isn’t always available so they compromise and try to extract it from other sources.

For example, they should use accurate traffic counts when assesing the need for road design changes and the construction of better facilities for cycling, either cycle lanes or separated cycle paths.

The trouble is, they sometimes rely on those induction loops embedded in the asphalt and that are often there as part of the traffic signal system. They believe they detect every wheel that crosses them. They are so, so wrong.

While it’s common for the loops to ignore cyclists altogether, it seems from research at Ohio State University that they can’t always detect the lumps of metal that are cars and trucks.

Some of the induction loop counters were wrong by a massive 52%. Such inaccurate data must never be used in designing cycling facilities on or adjacent to the highway.

Head case

There are three things that careful researchers avoid:

1. Entering the febrile arena of discussion about bicycle helmets

2. Questioning head on accepted wisdom, such as that which evolves from Cochrane reviews

3. Taking the time to correct the mistakes made by others.

So three cheers to Rune Elvik at the Institute of Transport Economics in Norway and editor of Accident Analysis and Prevention.

He’s done all three in a dense little paper, which will infurate the pro-helmet lobby because one of its conclusions is that “no overall effect of bicycle helmets could be found when injuries to head, face or neck are considered as a whole.

Zombie stats and ghost papers


It seems fitting to tell a little scary story as Halloween approaches.

They’re called zombie statistics – figures that should be dead and buried because they have no substance but they keep coming back to life.

Most of them don’t matter and are merely symptoms of everyone’s desire to believe in powerful myths.

But every so often there are zombie stats that are used to endorse serious messages and that’s just not right.

I came across one when trawling for data about helmets. The statistics page of the Bicycle Helmet Safety Institute cited figures published in a Children’s Safety Network pamphlet along with a link to download it.

One stood out. It said that “Head injuries accounted for 62.6 percent of bicycle fatalities” among children and youth age 0 to 19 in 2000 in the US.Zombie statistic in Children's Safety Network leaflet of 2008

It was clear, unequivocal and shocking.

I downloaded the pamphlet but was frustrated because it showed the statistic came from an unpublished paper.

Is this a real paper or a ghost?

So I contacted the author of the “unpublished paper”, Eduard Zaloshnja.

Eduard replied promptly and disarmingly honestly – sayng that no such paper existed, only a set of data he’d produced for the Children’s Safety Network.

So I asked for a copy of the data set but Eduard had lost it some years before when his hard drive crashed. Hmm, the origins of the startling figure showing the awful truth about head injuries killing a nation’s young cyclists was proving elusive.

Of course, it was reasonable to assume that the Children’s Safety Network, whch had commssioned Eduard’s research, would have a back-up of his data so I asked for a copy.

Two weeks passed.

Then the Library Assistant in the Network’s Education Development Center wrote to say they didn’t have a copy “since the paper is not yet published” and she apologised for the inconvenience.

Within two more weeks the Children’s Safety Network published a new, expanded, brochure and you can still download it here if you’re quick. The zombie statistic seemed to have been vanquished – at least, it doesn’t make an appearance.

Yet, eerily, the new pamphlet still contains a reference to Eduard’s unpublished paper, even though he says he never wrote such a paper and created only data which, unfortunately, disappeared several years ago when his computer crashed. How strange.

I guess the repeated reference to Eduard’s ghost paper might simply be a proof-reader’s error. More worrying, though, is the continued publishing of the zombie statistic on the Bicycle Helmet Safety Institute’s page.

They were told more than a year ago that, as neither Eduard nor the Children’s Safety Network could confirm the frightening data, it would be sensible to remove it from their Statistics page.

Spookily, it’s still there and it’s still a little scary – not because it’s true but because it can’t be substantiated. I hope it goes away soon.