Anti-Buzz: Malaysia Airlines Flight 370

by Andrew Emmott on October 4, 2014

in Anti-Buzz

newface-620x461Andrew has been writing Anti Buzz for 4 years resulting in almost 200 articles. For the next several weeks we will revisit some of these just in case you missed it.

Last week I discussed probability in regards to something fun – the NCAA basketball tournament – but there’s another current news item that is thrusting statistics and probability into the mainstream, albeit a much more somber one: Malaysia Airlines Flight 370, which vanished mysteriously and as of this writing still evades discovery.

When the story first broke, the old needle-in-haystack metaphor was trotted out, with the additional complication of “We’re not even sure which haystack to look in.” In this case the haystack metaphor is more apt than you might guess. Consider a flattened map of the surface of the earth, and then consider that you add hills to the map, not by observing where the actual real hills are, but by raising the altitude in places where you thought it was more likely for the missing plane to be. The visualization might look sort of like this:

So when the search first began statisticians, given what they knew at the time and which “probability map” they could draw, were more or less staring at a bunch of haystacks, with no good idea of where to begin.

You’ve seen these sort of probability maps before, even if you don’t realize it, unless you’ve never looked at a bell curve before. These curves are representative of one type of probability density function; you’ll notice it sort of makes a “hill” over one location. This is the one dimensional version of what I’m talking about; one axis of possibilities, with altitude adjusted for likelihood.

As more information has come out, they have reduced the number of haystacks, as well as moved them around. As of this writing, they have just moved the search area to a specific patch of Indian Ocean, off the west coast of Australia.

Remember what I said last week about the most likely thing being itself unlikely. I’m sure some statistician out there has a model of the situation and could point to an /exact/ spot on the Indian Ocean and say, “This is the most likely place for the plane to be,” and they could even be right, (about it being most likely), and yet the odds of the plane being exactly right there would be pretty long.

As with the NCAA bracket, you can improve your odds if you are allowed to select a region of possibilities. 15 million educated guesses at the NCAA bracket were still not even close to good enough to win, but statisticians can still map out a region of the ocean and make assertions about how likely the region as a whole is.

The more certainty you require, the larger the region becomes. I could, off the top of my head, draw a region that includes the missing plane with 100% certainty, and so could you; that region would be the entire surface of the earth. The fact that somebody has narrowed the search down to “probably in this big area” is actually pretty amazing.

The other tech story coming out of this tragedy, the one that maybe hits home a little harder, is how, in this age of smart phones and GPS and cloud storage, can a plane get lost like this in the first place? I’m not qualified to answer that question, other than to say that “always on” tracking of every plane ever even when something goes wrong like it did here sounds to me like an infrastructural nightmare, and perhaps not even worth the expense. However, less excusable, is the fact that the flight records are trapped on this lost plane. Bandwidth for planes is expensive, but it seems to me that “the cloud” should be able to grab hold of at least a partial back up of flight records and keep them here on terra firma so that we don’t have to actually find the black box at the bottom of the ocean before we can ask it how it got there.

by: at .


Comments on this entry are closed.

Previous post:

Next post: