On Wed, 2011-03-16 at 19:18 -0700, RonaTD wrote: > > On Mar 16, 11:10 am, Jan Heine <hein...@earthlink.net> wrote: > > At Bicycle Quarterly, we are considering looking at the statistics > > and figuring out whether helmets make riding safer, whether risk > > compensation really is a factor, etc > > As somebody who makes his living being dangerous with statistics, let > me gently wade in here and suggest that statistics are useless for the > purpose of helping an individual person make a decision whether to > wear a helmet. Whatever statistics you come up with will be averages > across all sorts of controlled and uncontrolled factors. Lots of > conditional probabilities get the conditions rubbed away and become > seemingly unconditional. > > The reality for an individual cyclist, at an individual point in time > and space, is that things are extremely conditional. And, when you > start piling on all those specific conditions, your statistics are > meaningless. > > In the language of insurers, we're dealing with problems of incidence > and severity. For most people who wear helmets when they ride, it's a > pretty simple calculus. The cost of wearing a helmet is extremely > small, so incidence isn't much of a factor. The potential severity of > injury from a fall, and the perceived reduction in likely severity if > there's a fall, makes the perceived benefit of wearing a helmet much > greater than the cost. The problem with a brainless "expected loss = > incidence * severity" analysis is that it only works for an insurance > company. For an individual, the equation becomes much closer to > binary. The expected severity is high if there's a fall (especially > when high velocity and/or high mass moving objects are involved), and > it's a conditional calculation: no fall, no severity; fall, high > severity. Individuals don't get to benefit from the law of large > numbers to even out the severity distribution, especially because it > only takes one incident to put the individual in the right-censored > part of the distribution. > > So, if you want to add something meaningful, I suggest focusing on > severity. Collect all the data you can about cycling accidents, with > and without helmet use, and measure the distribution of injury type > and severity, controlling for whatever you think is relevant. Forget > about trying to prove anything about incidence, that's a fool's > errand.
Bravo! And when you're thinking about "severity" let's not limit the field to brain injuries alone. I'll bet if she saw the scrapes on the side of my helmet that time I went down sideways going around the traffic circle at Mount Vernon, when my wheel got caught between the two concrete lanes, and imagined them on the side of her face, Anne would have rated that as a "pretty damn severe" injury. -- You received this message because you are subscribed to the Google Groups "RBW Owners Bunch" group. To post to this group, send email to rbw-owners-bunch@googlegroups.com. To unsubscribe from this group, send email to rbw-owners-bunch+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/rbw-owners-bunch?hl=en.
