Died: 10 October 1961 in Pittsburgh, Allegheny, Pennsylvania, USA

**Edwin Olds**' parents were Edwin Nelson Olds (born 18 November 1872 in Wheatfields, Niagara County, New York), a farmer at Pendleton, New York, and Effie Ruth Wells (born 11 September 1876 in Mapleton, Niagara County, New York). Edwin Nelson and Effie Ruth were married on 14 April 1897 and Edwin, the subject of this biography, was their first child. In fact Edwin was ten years old before another child was born, namely Ada Ruth born in Mapleton, Niagara County. The family was not completed until 1915 when Doris, a second sister for Edwin, was born.

Olds studied for a B.A. at Cornell University in Ithaca, New York. He married fellow student Marion McNeil Knowles (born 21 Apr 1897 in Lockport, Niagara County, New York) in Ithaca on 29 March 1918. They both graduated from Cornell University in 1920; they had two children, David McNeil and Marcia Elisabeth. Olds undertook graduate studies at the University of Pittsburgh and was awarded a Master's degree. Appointed to the Carnegie Institute of Technology, he worked there for forty years until his death. Today the Carnegie Institute of Technology is part of Carnegie Mellon University but this was only created with the merger of the Carnegie Institute and the Mellon Institute a few years after Olds' death. He was an assistant professor of mathematics for many years but, after the award of a Ph.D. from the University of Pittsburgh in June 1931, he was promoted to the rank of associate professor of mathematics at the Carnegie Institute. Olds' main interests were in statistics and he was highly influential in directing some outstanding students into that area. In particular, in [2] (see also [1]) Frederick Mosteller describes how, as an undergraduate at the Carnegie Institute of Technology, he was influenced by Olds. Mosteller also gives us a good picture of Olds and the research he was undertaking:-

A little statistics and probability entered the physical-measurements course, and somewhere along the way we were asked to compute the probability of casting a total of9and of casting a total of10using three ordinary dice rather than two. ... although most students found the answers, the problem troubled me. When the class discussed this problem, I said to Dr Pugh, who was excellent at keeping us motivated and moving an extra step, "Most of us got the answer mainly by counting on our fingers. But, if you had asked about6dice or15dice, we'd still be counting. Is there a better way to do it for larger problems?" Pugh's greatness as a teacher came through. He said immediately, "I don't know how, but I think I know a man who might, Dr Olds. Why don't you ask him about it?"

Among the mathematics professors at Carnegie, Dr E G Olds was one of the few doing original research at that time. He was also teaching me calculus, and he had a huge voice, was very good humoured, and had astonishingly big teeth. He reminded me of other loud voices in my life, and I did not follow Dr Pugh's advice, though I saw Dr Olds nearly every day. One day soon after, I was working in the silent library when Dr Olds came in and shouted in his usual conversational tone of voice, "I hear you want to know about the dice problem. You must come see me some time about it." Me, whispering, "Yes, one of these days." He then said, "Well, you're not doing anything important right now. Come along," and led me off to his lair.

After a little, I got over my embarrassment and began thinking, and he began showing me slowly and carefully how to do the three-dice problem. ... Although I had loved mathematics all along, this was the first time I ever felt that I'd been working with a peashooter when I could have had a cannon, and furthermore, that the tricks I had learned could produce such a marvel. I thanked him sincerely and got ready to hurry off to class, but before I reached the door, he said, "Just a moment, young man. Come back." He quickly pulled a book off the shelf, turned to the end of the first chapter and marked off ten problems. "If you liked the probability problem and the generating function, you'll like this. Bring the problems back in two weeks." Authority meant a lot then, and so I did. Each week he'd check off problems in another chapter, book after book.

I complained to my mother that I was taking an extra course and not getting credit for it, and why did I have to do it? "I don't know, dear, but I'm sure he wouldn't ask you to do it if it weren't good for you." Soon I was a mathematics rather than a physics major. If week after week, year after year, and book after book, you study a chapter and solve ten problems in a subject, it's hard not to have some feeling for it and investment in it. Furthermore, I could see how it could help my poker playing. When I mentioned this to Olds, he seemed uneasy but did not reprove. He was a serious churchman. Despite this, one summer I wrote him from the road job about a poker problem I had found instructive and how I had gotten an approximate answer. He responded with a long letter bringing up a point I had missed, though it luckily didn't matter a lot, and showing a way to do the problem exactly. I believe his own doctoral dissertation was on probabilities in bridge. And he was a bridge player.

In part of his research, he was engaged in finding the exact distribution of the rank correlation coefficient for samples of size7and8. This led to enormous piles of paper on each of which a substantial array of numbers were written - oh, for today's high-speed computer! The hitch was that he could tell from looking at his answer that there were some mistakes in the counts, more than one. Some governmental funds were available for students to earn by helping professors do research, and so he unleashed me on this pile to check the answers. I did this by merely redoing his calculations, a method that for various reasons is not likely to be a good check. ...After a while, I began noticing something of a pattern - something familiar page after page. ... I quit doing the actual work and began playing around with the numbers, and sure enough, the totals always provided a special pattern. If they didn't, then something was wrong. Now I had a theorem, could I prove it? As often happens when you know the answer, the proof comes rather easily.

When I showed Dr Olds the result, he was clearly pleased. He also was able to turn the task of checking over to a clerk because now all that had to be done was a series of additions to get a definite check. And the new check was much faster and more reliable than the repeated calculation. I suppose this was my first research experience in mathematics. It did use some of the mathematics I had learned in my special readings assigned by Dr Olds, and I probably could not have done it with the equipment I picked up in my regular course work. Most regular work emphasized calculus or geometry; the discrete mathematics I learned with the outside readings was what was required.

During World War II, Olds served as chief statistical consultant to the Office of Production Research and Development of the War Production Board and also taught a course *Quality Control by Statistical Methods* for the War Production Board at Ohio State University. This intensive training course was designed to produce graduates with skills useful for the United States' war effort. In fact many graduates from Olds' course went on to become founding members of the American Society for Quality.

Olds was passionately interested in teaching as well as in research. He published over fifty articles, both research papers and articles on the teaching of statistics and mathematics. Examples of his research papers are *Acceptance Sampling By Variables* (1947), *The 5% significance levels for sums of squares of rank differences and a correction* (1949), *A note on the convolution of uniform distributions* (1952), and (with Norman Severo) *A comparison of tests on the mean of a logarithmico-normal distribution with known variance* (1955). Here is an extract from the introduction to the 1955 paper:-

The present investigation is concerned with the application of the logarithmic transformation to the problem of testing an hypothesis on the mean of a logarithmico-normal variate with known variance. An experimenter can fail to recognize the need for a transformation and simply proceed to apply normal theory tests to the original data; or he can properly transform the data and then apply a normal theory test to a parameter of the transformed scale. Each of these testing procedures is investigated in detail. Finally a third test procedure is developed by using the Neyman-Pearson Lemma for testing simple hypotheses. A comparison of these tests is then made by means of their operating characteristics and some asymptotic properties obtained. It is found that the three procedures give quite different results unless the mean under the null hypothesis is large relative to the standard deviation.

Examples of his papers on teaching are *A Fresh Start* (1938), *Why Learn Mathematics?* (1939), *We Discover the Meaning of Curvature* (1940), *Using the Experimental Approach in the Teaching of Statistics* (1954).

Olds was honoured widely for his important contributions. He was Vice President of the National Council of Teachers of Mathematics in 1943-45. He was president of the Institute of Mathematical Statistics (1954). The American Society for Quality awarded him their Brumbaugh Award in 1953 and, in the following year, their Shewhart Medal, named in honour of Walter A Shewhart, for:-

... outstanding technical leadership in the field of modern quality control, especially through the development to its theory, principles, and techniques.

He died in his Pittsburgh home, aged 61, from a heart attack.

**Article by:** *J J O'Connor* and *E F Robertson*

**July 2011**

[http://www-history.mcs.st-andrews.ac.uk/Biographies/Olds.html]