4 ms·

sampling bias debate reminds me of this stats legend from ww2: the USAF wanted to armor planes where the bullet holes were. statistician abe wald recommended a

sampling bias debate reminds me of this stats legend from ww2: the USAF wanted to armor planes where the bullet holes were. statistician abe wald recommended armoring where the bullet holes weren't on the theory that those were the non-survivable hits can't interpret a statistic until you think about what's excluded from the sample

jldugger 7y ago
We don't really know that 'the USAF' wanted anything; that bit has the 'citation needed' flag set in wikipedia. The paper itself makes no comment on the presumed answer to the question, focusing mostly on mathematical models. It certainly makes for a good drama though.
pirocks 7y ago
When I first heard this story it was the RAF not the USAF. I somewhat doubt how true it was.
- awinter-py 7y ago
  yup, 'stats legend', not history there's an american math society article about this -- it evaluates the fun version of the story in the context of wald's actual paper and a journal entry by another scientist named wallis
opo 7y ago
>sampling bias debate reminds me of this stats legend from ww2: the USAF wanted to armor planes where the bullet holes were. statistician abe wald recommended armoring where the bullet holes weren't on the theory that those were the non-survivable hits It wasn't quite that simple, but the story has a gem of truth. As this article quotes Stephen Stigler (son of George Stigler who worked with Abraham Wald): >..."The military was inclined to provide protection for those parts that on returning planes showed the most hits. Wald assumed, on good evidence, that hits in combat were uniformly distributed over the planes. It follows that hits on the more vulnerable parts were less likely to be found on returning planes than hits on the less vulnerable parts, since planes receiving hits on the more vulnerable parts were less likely to return to provide data. From these premises, he devised methods for estimating vulnerability of various parts." http://www.ams.org/publicoutreach/feature-column/fc-2016-06 http://www.ams.org/publicoutreach/feature-column/fc-2016-06 This paper goes into detail for those interested in statistics: https://people.ucsc.edu/~msmangel/Wald.pdf https://people.ucsc.edu/~msmangel/Wald.pdf

jldugger 7y ago

We don't really know that 'the USAF' wanted anything; that bit has the 'citation needed' flag set in wikipedia. The paper itself makes no comment on the presumed answer to the question, focusing mostly on mathematical models. It certainly makes for a good drama though.

pirocks 7y ago

When I first heard this story it was the RAF not the USAF. I somewhat doubt how true it was.

awinter-py 7y ago
yup, 'stats legend', not history there's an american math society article about this -- it evaluates the fun version of the story in the context of wald's actual paper and a journal entry by another scientist named wallis

opo 7y ago

>sampling bias debate reminds me of this stats legend from ww2: the USAF wanted to armor planes where the bullet holes were. statistician abe wald recommended armoring where the bullet holes weren't on the theory that those were the non-survivable hits It wasn't quite that simple, but the story has a gem of truth. As this article quotes Stephen Stigler (son of George Stigler who worked with Abraham Wald): >..."The military was inclined to provide protection for those parts that on returning planes showed the most hits. Wald assumed, on good evidence, that hits in combat were uniformly distributed over the planes. It follows that hits on the more vulnerable parts were less likely to be found on returning planes than hits on the less vulnerable parts, since planes receiving hits on the more vulnerable parts were less likely to return to provide data. From these premises, he devised methods for estimating vulnerability of various parts." http://www.ams.org/publicoutreach/feature-column/fc-2016-06 http://www.ams.org/publicoutreach/feature-column/fc-2016-06 This paper goes into detail for those interested in statistics: https://people.ucsc.edu/~msmangel/Wald.pdf https://people.ucsc.edu/~msmangel/Wald.pdf