Wednesday, 28 July 2010

Speaking of Statistics

Someone send OPP Chief Julian Fantino back to school (and send the reporters covering this as well).

These are irrelevant statistics:

In 2008, there were 322 highway traffic deaths on OPP-patrolled roads compared to 451 in 2007, the release states. Last year, there were 310 highway traffic deaths.

How many drivers were on the roads? Important context to the whole discussion no?  Especially when there are approximately 9 MILLION licensed drivers in Ontario.

This is how you report statistics:

In 2006, Ontario was the North American leaders in road safety (0.87 fatalities per 10,000 licensed drivers)

See that PER 10,000 licensed drivers? That's what us logical people like to call "context".  It frames the discussion and is critical to discussing trends.  (A better measure would be per 10,000 cars on the road or per X kilometers driven.  But licensed drivers gives us some sort of meaningful context.) (And yeah, on this same page I linked to they make the same mistake of reporting an unquantified number of fatalities.)

So a change from 310 to 451 out of 9,000,000 is within statistical error is it not?  Yes every death is a tragedy, but unless you can show a clear trend, we really shouldn't get worked up over this now should we?

And in case we're still not clear, here's a hypothetical.

In 2006 100 students out of 1000 at Main St. High School graduated with a straight A average.

In 2007 102 students out of 1500 at Main St. High School graduated with a straight A average.

If you were the principal, would you be bragging that 2007 showed an "improvement"?  If you answered yes, time for you to go back to school.

No comments: