This article gives a great insight into the nature of randomness:
http://www.wired.com/2012/12/what-does-randomness-look-like/
I've always argued that clusters of plane accidents are reassuring signs that they truly are random. If plane crashes happened with precise, predictable regularity, once every three weeks, say, then that would be a sign that there was some sort of terrible uncontrollable force at work in the world.
This paper argues it's a poisson distribution:
Time-evolving distribution of time lags between commercial airline disasters[1]
while others disagree:
In the case of plane accidents, the authors of Ref. 7[1] found that the time lag between commercial airline disasters and their occurrence frequency could be well described by time-dependent Poisson events. On the other hand, authors of Ref. 8[3] have found that beyond certain timescales the time dynamics of both plane and car accidents are not Poissonian but instead long-range correlated.[2]
[1]http://arxiv.org/abs/physics/0509092
[2]http://arxiv.org/abs/0901.3183
[3]http://www.sciencedirect.com/science/article/pii/S0378437107...
I feel it's important to point out that this is not a cluster of three airliner accidents. It's a cluster of two accidents and one crash due to military action. While they presumably didn't intend to shoot down an airliner, the only accidental part was that they killed different people from who they wanted to kill.
I don't think this affects the probability discussion much, but it's good to call it what it is.
> In fact, Ranter says it is more common for an accident to happen just one day after another crash than two, three or more days later.
As written, a poisson distribution wouldn't explain this. If you have a crash on day 0, then a crash on day 1 is no more likely than a crash on days 2, 3, etc.
Day 1 is the day most likely to have the next crash, but it's no more likely to have any crash.
(It could easily be the case that Ranter actually found "...just one day after the previous crash than two, three or more days later", and this subtlety got lost somewhere down the line.)
I don't agree with the term 'accident' for crashes because they can always be attributed to causes, not just chance.
This is even more true when missiles are involved.
Yesterday in Toronto, a plane made an emergency landing. The news ignored it as just an usual everyday thing, because you know, it happens everyday... Just kidding, they actually called it a "multiplication" of aviation incidents.
"But the chance the next crash is on 3 August is (364/365) x (1/365), because the next crash occurs on 3 August only if there is no crash on 2 August."
Why is a crash on 3 August dependent on there being no crash on 2 August? Surely there could be crashes on both 2 August and 3 August.
People pay more attention to airplane accidents right after a large accident event.. Google plane accidents and you will see that they happen all the time, they just don't get the world wide news coverage.
My question is how far has decades of cut throat competition eroded the safety and maintenance standards of all airlines?
Minimum standards are high - and well respected. But if you shave off costs - even if each one is trivial - put back investment, dial back on the training, will I have a cumulative effect?
In short, when will air travel trend towards rail and road for accidents ?
This is absolutely correct and insightful. Glad to see it on the BBC.
Now here's a scary question: when you want to give someone a bonus because of their successful performance over the last 6 months, how likely is it that their string success is a coincidence?
You're forced to make a similar conclusion, and if you don't, think about why you have the bias you do.
All crashes as mentioned:
- MH 17 http://www.aeroinside.com/item/4365/malaysia-b772-near-donet...
- Transasia in Taiwan http://www.aeroinside.com/item/4389/transasia-at72-at-makung...
- Swiftair MD 83 in Africa http://www.aeroinside.com/item/4393/swiftair-md83-over-mali-...
I read a book once - I think it was the Gift of Fear (but I might be wrong) - that said that plane crashes often occur in clusters, same with other tragedies like mass shootings at schools etc. The author suggested this was because people who might do those things - or in the case of pilots, pilots who might want to commit suicide and take the plane down with them - get prompted to act when they see other incidents have occurred. So according to that author, these events are not random.
Obviously a civilian plane being shot down would lie outside this theory (he suggested more plane crashes were due to pilot depression than mechanical fault/other factors).
"the most likely maximum number of crashes of commercial planes with over 18 passengers in any eight-day window over 10 years is exactly three".
So if another crash occurs are we allowed start the conspiracy theories?
Time between disasters has exponential distribution (or the number of disasters in given time adheres to Poisson distribution). Something we were taught at the university.
For those interested, a good source for a lot of publicly reported crashes, accidents, incidents and whatnot is The Aviation Herald. Here's the list of all their recorded crashes sorted by occurence date (toggle the icons next to "Filter" to see more kinds of events): http://avherald.com/h?list=&opt=7681
The Werther effect could be another explanation for clusters of airplane crashes:
https://fc.deltasd.bc.ca/~dmatthews/FOV2-00074762/S02DB0598....
The is essentially the Texas Sharpshooter fallacy.
While it is impossible to reduce chaotic systems to individual factors, and it is likewise impossible to predict the effect of an individual factor upon the system as a whole, it is certainly possible for one factor with a global effect to move the entire system towards or away from a probable outcome.
If we postulate that, now absent the selection pressures that have shaped human intelligence over the last few million years or so, human intelligence is likely to decline, then we can ask ourselves where this decline might be likely to first show up in the chaotic system of human endeavour.
One possible answer is that it will appear first at the boundary layers: the places where a critical level of human intelligence is required to keep a complicated task operating.
I propose that flying passenger aircraft is such a task. A critical level of intelligence must be maintained by a very large number of people in order to keep passenger aircraft in the air. Everyone, from designers to manufacturers, to QC, to maintenance to pilots to airline management has to function above a certain critical level to perpetuate the activity.
It is possible that clusters of aircraft accidents are purely random and part of the complex system that is air travel. However, it is possible that clusters of aviation accidents represent crossings of the boundary layer resulting from the change in a global factor, like human intelligence, that has moved the entire system probabilistically.
The details of some recent accidents should give us pause. The series of over-control/mis-control accidents including AF447, Colgan Air and others defy reasonable explanation, and they appear to have no precedent in recent passenger aviation. MH370 and MH17, so far as we can see, have no reasonable explanation other than unaccountable human behaviour (failing to communicate over the course of seven hours flying in the case of MH370, and navigating over a war zone in the case of MH17).
It is possible, although certainly not provable at this point, that we are simply becoming too stupid (in general) to fly passenger aircraft safely. It may be time to switch to fully automated aircraft systems.
I'm pretty happy someone actually wrote something about this, I was surprised I didn't hear any news outlets talking about the connection beyond just mentioning the other accident.
I'm also happy to see some at least slightly non-trivial statistics in the mainstream.
One glaring point though: the liklihood of another crash might not be high given whatever statistics, but those don't reflect the fact that someone shot a missile at one of those planes. I might risk a flight in Africa or Taiwan or where-ever, but you won't see me flying anywhere near Russia/Ukraine anytime soon even though people obviously thought this was a perfectly reasonable thing to do.