[kqr]

> The problem with lists of major accidents is they aren't necessarily a predictor of future accidents, and in some cases may be the reverse. An airline suffering a major accident is likely to lead to a significant safety overhaul.

This should be a somewhat testable proposition, right? Does time since accident increase or decrease the probability of next accident?

(Practically, plotting the distribution of inter-accident durations and seeing if the tail falls off exponentially, slower, or faster.)