I deal with statistical analysis quite a lot. One of the problems with using it is that people so frequently insist that the lessons of the past no longer apply.

For example, engineers will assert that a specific class of failure modes no longer apply because “we fixed that and it will never happen again.” Therefore, they argue, the probability of failure must be less than the historical record indicates.

Unfortunately, there are other things they have not fixed that will surely rear their ugly heads. And there is no evidence that any type of problem is ever truly fixed. Memories fade and other urgencies arise to displace older priorities.

In May 1986 a Delta booster broke up after launch from Cape Canaveral. The cause was a chafed wire and the entire space launch industry underwent something of an epiphany in regards to that type of problem. So it was fixed, right?

In August 1998 a Titan IV booster broke up after launch from the Cape. The problem was once again a chafed wire. And just a few years later the Titan engineers argued they had fixed their vehicle and that past failures should not be “counted against them.”