I am sad to report that Carl Bennett passed away in early June at 92. Carl played an important role in the development of statistics and its practical uses in the twentieth century.
Carl was a PhD candidate at the University of Michigan when the Second World War broke out. He was sent to Oak Ridge where he was the first statistician on the Manhattan Project. When his work at Oak Ridge was completed he went back to Michigan to finish his studies, but it was not to be. Edward Teller needed Carl at Hanford for the hydrogen bomb project. Carl finally completed his doctorate long distance from Hanford.
When that project was completed the US Navy asked Carl to write a book about the use of Statistics in industry. The Navy funded the project. Carl moved to Princeton and with Norman Franklin wrote the book. “Statistical Analysis in Chemistry and the Chemical Industry” was published by John Wiley & Sons in 1954. Several generations of statisticians know it as Bennett & Franklin. Hanford was operated by General Electric under contract to the US government. Carl was GE’s Chief Statistician.
By the late 1950's the Government understood how harmful radioactive fallout from nuclear tests could be. President Eisenhower initiated top secret talks with the Russians to ban atmospheric testing. Ike told Carl that he knew the Russians would try to cheat so he did not dare sign the treaty until he had a way to verify compliance. He asked Carl to work out a way to do that. The United Nations Special Commission (UNSCOM) grew out of that effort. Carl moved to Vienna to be at UNSCOM's headquarters. After retirement Carl returned to Vienna every year to attend the annual meeting of UNSCOM and to check on the quality of the Austrian Heurige (wine bars).
Carl returned to Hanford. The contract to manage Hanford was awarded to Battelle and eventually Carl moved to a Battelle research facility near Seattle. I first met Carl at an ASA meeting at the University of Washington. My mentor had instructed me that there were many textbooks on Statistics, some good and some not so good but that if I relied on Bennett & Franklin I wouldn't go wrong, so when I realized who Carl was I was star struck. I asked what he was doing and he said he was retired. I asked if he would be willing to do some work for my company. Fortunately, he knew who we were and he agreed.
We often use full factorial experiments in our industrial work. We evaluated the results using Yates' algorithm. Sometimes the important result was a single interaction. But whenever that was the case Yates' algorithm gave the wrong answer. If you put the experimental results into the original experimental matrix you could see the correct result but you couldn't calculate it. Carl said it came from the way Fisher and Yates thought of interactions. My most memorable day with Carl was spent in our office shouting at each other and running up to a white board to scribble things on it. The result was Bennett's Method, a way to calculate the result of a full factorial experiment correctly.
We miss you Carl.