Originally Posted by
RazrSkutr
... I may be 1000 times more likely to die when being hit by an asteroid than when hit by an automobile, but presenting that as "you are disproportionately more likely to die when hit by an asteroid than a car" does not mean I should start looking for an asteroid deflection shield.
In this case, the asteroid example is not a valid comparison; after all both sets of data show about a third of fatal motor vehicle - bike crashes are overtaking. And per below, in a rural cycling environment they can be common enough they are the ones to watch out for at the times you are not crossing an intersection or driveway.
Originally Posted by
RazrSkutr
And I think we can make a good guesstimate that we are overtaken far more often than we pass cars at intersections.
I agree. However, when you are not crossing an intersection or driveway in a rural setting, what situations should you be wary of? I would say the data shows overtaking motorists are one of the prime situations to be on the lookout for compared to, oh, say, asteroids (not even the CPSC would claim helmets offer protection from those!).
Originally Posted by
RazrSkutr
In any event the other thing that your data (incomplete though it is) show are:
1. As a percentage of total incidents ... e.g. as is indicated by other data things have gotten a lot better since the 70s and there's no obviously massive threat from overtaking incidents.
One of the tasks I do at work deals with count data. Administrators are wont to get all in a tizzy about some statistic that is expressed as a percent when the underlying raw count data is so small that the calculated percent has a very wide confidence interval and no statistically significant change. The statistics of rates and proportions (to use the title of one of the classic textbooks in this area) are such that you need surprisingly large counts or lots of time periods to be able to say much.
Comparing the count data in the 1977 Fischer & Cross study vs the Bike Crashes of the 90s data using percentages is one such example. Lets do one of the most common statistical tests on 2X2 tables of these data to see if there is really any nonrandom change in the proportion of motorist overtaking crashes.
First, the Nonfatal MV crashes:
Nonfatal Overtaki.gif
Not even close to a significant change in the proportion.
Then, the fatal crashes:
Fatal Overtaking &.gif
the p-value again does also not even get close to the 0.1 or 0.05 range.
There are more sophisticated tests to deal with multiple 2X2 tables at once or Rx2 tables, but at least the one I tried failed to detect a nonrandom change over time in these data.
Finally, at least in North Carolina, there is the fairly large amount of data collected over recent years showing an increasing proportion of both rural and urban crashes due to motorist overtaking. This despite a stable to slightly decreased total number of crashes. Some cyclists from NC thought there were several plausible contributing causes, including patterns of development where new developments are linked by relatively narrow rural roads, more wide vehicles such as SUV's and, of course, distracted driving in its many forms:
Motor Overtake Bike Crash..jpg
In conclusion, particularly in a rural setting when not crossing intersections or driveways, I keep an ear and eye for what is coming up behind me plus the other aforementioned steps. The recent motorist overtaking crashes that left a
Spanish cyclist in RAAM paraplegic,
killed an experienced rider in NC, or
killed Paige Hicks on her second cross-country charity ride are, sadly, not rare events.