Wednesday, April 25, 2012

I've just found another person for my 'Nerds at War' series (a general interest book that I'll get round to writing one day).

Abraham Wald
Abaham Wald was born in 1902 in what was then Austria-Hungary. His hometown is now part of Romania, but he belonged to an earlier map of Europe, in which you could be German or Jewish (he was Jewish) but hail from just about anywhere in eastern Europe.
He studied mathematics at the University of Vienna, which even before the Nazis arrived wasn't the most Jew-friendly of institutions. One of his supervisors, however, managed to get him a position teaching in the school of economics, where he remained until 1938. When the anschluss came, Wald was already putting the finishing touches on his escape plan, and he was gone long before the knock on the door came.
He ended up in the United States, working at Columbia university, where he was was finally able to work on research for publication (something that he had been previously barred from doing). He specialised in the more complex and abstract branches of statistics, though he also undertook some economic analysis work for the government when the war began.
His most notable contribution to the war came in 1944, when the US Air Force asked him to look over their damage statistics. The strategic bombing campaigns of the European theater were exacting a heavy toll on aircraft and airmen, with the airforce losing hundreds of bombers to anti-aicraft fire every week. They were always looking for ways to improve their aircraft, and to this end they collected vast reams of information from the mechanics who fixed the planes when they got back to base. They had diagrams showing where planes got hit most often, and had resolved to add extra armour plating to those vulnerable spots. To make sure they were not making any mistakes they showed their statistics to Wald.
Wald got back to them a few days later with some surprising news. They were, he explained, looking at the statistics the wrong way round. The data set they had was corrupted by what statisticians call 'selection bias' -- it did not show the damage sustained by all their planes, but  rather the damage sustained by those planes that made it back to base to be repaired. In other words, what the diagrams and statistics showed were not the aircraft's most vulnerable points, but its least vulnerable points, places that could be hit again and again without bringing the plane down.
Wald produced a report that was a mirror image of the previous one, the spots he highlighted as being in need of extra armour were those where the air force had few, if any records of aircraft being hit. Wald saw that these were not places where no plane ever got hit, they were places where no plane ever got hit and survived. 
History does not record what came of his insight, but even if his work only marginally improved the survivability of a US bomber crew, then it would have saved hundreds of crewmen's lives.