I told you so...

August 19, 2014 9:45 am

This report from Statista has been making the rounds telling everyone how Honda Accords and Civics are magnets for theft.

Now it is important to remember our discussion about Bayesian statistics and why the headline of the article and chart, "Honda Owners Are Most At Risk From Car Theft," is simply wrong.  This data makes no attempt to account for the number of these vehicles in use when considering how likely a vehicle is to be stolen.  What this data really says is, "Of cars that have been stolen, we counted more Honda Accords and Civics than other vehicles."  Which is slightly different and explicitly says nothing about how likely Accords and Civics are to get stolen.  Claiming that, from this data, Accords and Civics are most likely to be stolen is disingenuous at best.  You must include the "prior" that accounts for the number of Accords and Civics that exist in the first place in order to say anything reasonable about the likelihood of Accords and Civics getting stolen.

If you don't at least know about Bayesian statistics you'll be misled by people who, either intentionally or not, report misleading data.  And I guarantee this report is being used today to sell additional insurance coverage to someone sitting in a Honda dealer buying a car (the same way they tried to do so to me in 2007).

Leave a Reply

Your email address will not be published. Required fields are marked *