by lutorm » Wed Nov 22, 2017 1:34 am
That's of course good news, but my concern is that with only ~30 fatal accidents per year, there's a 1/3 chance that the number would be outside of the 21-39 range on any given year just from random variations, without any underlying real changes. We should be careful of claiming changes like these as causal since there's a large amount of random variation and when we get a random fluctuation to larger number of fatalities on some year people will claim that things are "going in the wrong direction" when it's really just noise.