Remembering Jimmie Savage

Leonard, Jimmy Savage. I feel like I don't know him. Can I call him Jimmy Savage? Should I call him lj? What, what, what, what was. Tell us about Jimmy.

And I'll, I'll get into that because it matters, it really leads to things, uh, in his, uh, uh, adult life and his attitude. But the big thing with him was his eyesight. And he was born with a congenital defect. Um, and it was complicated by, uh, myopia. So that when he read something, he had to take it up to his eyes like this, and he turned, and his eyeballs were continually moving.

Uh, it was a very sad thing that when he was born, back to how he got Jimmy when he was born, uh, his mother went through some bad. Uh, things associated with the childbirth and wasn't able to pay attention to things like, what's the name of the kid? Um, and so, uh, at one point a nurse was, you know, didn't have a name, so she wrote down a name. It was Jimmie with an IE and, um, when it came time to name him, um. There was a tentative name, and then the mother sort of picked the name Leonard.

Uh, and, but the, the nurse kept calling him Jimmie anyway, uh, it stuck and he liked it so that he was, it's his middle name, uh, but it was also his nickname and, uh, for reasons known only to him, he liked it. I, I suppose it was, you know, everybody called him that when he was a young person. And so he, he went by that name. But it was, uh, all his friends definitely called him Jimmy and his enemies, uh, probably called him Savage.

So I applied to various places, uh, and I got a fellowship at, uh, Yale and, and, uh, went to Yale. And I met Jimmy when I walked into his office 'cause he was. Uh, there, he and Frank Anscombe were the two big names. They had recently formed the department. Frank was the first chair, uh, and then Jimmy eventually became, uh, the chair of the department. But it was, uh, it was small. Uh, and so I met him. Uh, I took a course from him, uh, in the very first semester I was there.

Uh, it was a course out of feller and it was, um, a, a course that undergraduates could take, but it was also for graduate students. And so I'll, I'll get into that. So anyway, that's how I met him.

Obviously, bays and, uh, so bays is, uh, deserves to be called the father of, uh, Bayesian statistics. But Savage, really, Jimmy really. Made it for, uh, bayesians. I mean it provided for the first time throughout statistics, it provided for the first time a rigorous, uh, definition of what statistics, what statistics was that It was in fact a mathematical discipline. Uh, and he started out. Just a, a little bit of that.

He started out thinking that statistics, by the way, he was not a statistician. He got his degree in mathematics. He was working in and wanted to work in physics and chemistry. Chemistry turned out. You know, that was sort of unfortunate because he couldn't see, uh, and it was, uh, how he went into the lab. So eventually he really fell in love with statistics and the philosophy of statistics, you know, what did data mean? And to him, he was. He was a polymath. He knew everything.

He knew everything about everything. I mean, it's, it is just amazing and I'll give you a few examples of that. Um, but he, he knew, uh, about, um, uh, various EE economics, you mentioned economics.

We worked on some examples in science that of course he knew about the science and he brought me into it. Um, he, his childhood had some ops. Mostly they were downs. Uh, we know about his childhood mostly because of Richard. Savage, who was his brother, younger brother by eight years or so. Um, and so Jimmy was Richard's father. Um, uh, Jimmy's father was instrumental in making Jimmy's life good and was a, a, a big positive.

Um, uh, the rest of his childhood was pretty bad, but, uh, his parents bought him an encyclopedia when he was very young, and he read it and, and he, he remembered it. Uh, just a very simple thing about that. Uh, the main thing I want to tell you in a minute. Uh, so he one time said to me, he said. Uh, Don, uh, uh, Jean, his wife Jean, um, is writing a project on Finland and she's looking for celebrities in Finland. She can't find many celebrities. Do you know any celebrities in Finland?

Uh, from that they Finn and I said, I think Victor Borger is a, uh, fan. And he said, no, Victor Borger is. Danish. How did he know that? Do you know what vi Victor Berger does? You probably don't know. He was, he was, he was a celebrity. Uh, he was a pianist. He was a comedian. And he had a piano that joked, I mean, he worked with the piano and various jokes in the piano, and he was really clever and, uh, talented, uh, but not very well known.

And he appeared in the TV shows, like, uh, ed Sullivan show some, you know, uh, things like that. Uh, and, uh, but he wasn't really very well known. Then a little bit more known than he is today. Uh, so how did he, how did he know that he couldn't watch television? I mean, he couldn't see. Um, and his, uh, childhood, just, I mentioned Richard. Uh, Richard, uh, said, um, that, and I wrote this down. He was a brilliant child.

But he paid no attention to what was going on in school because he couldn't see what was going on in school. Um, and his teachers said, you can't go to college. I mean, she put in, they put in really negative things about him. According to Richard, uh, the prevailing wisdom in the school was that he was feeble-minded.

Uh, and I, I mumbled things about, you know, uh, um, what the probability was, and I was really interested in that separate from my own personal interest. You know what? How do you do this? How do you find the probability that the unborn is, is going to be a boy? And this was in the days PR prior to amniocentesis and any of these. And it used to be that people would say. Um, you know, pregnant woman would come into a room and would, would meet somebody and they'd say, oh, you're carrying low.

It must be a, it must be a boy. Uh, and you know, it was really these, uh, sort of old wives tales thing. But, so I was interested in, in, in, in pressing your mother with something that was, uh, uh, legitimate. And so I said. Uh, to Savage, I said it, it's silly to say, to use a maximum likelihood estimate, which is one I know it's not one. Uh, so how do I do this?

How do I find a prior distribution, let's say I'm willing to accept that, um, my wife and I have the same probability forever, uh, of, of a boy, but it varies potentially between Madrs. Um. And, uh, so let's assume exchange ability. Um, and, uh, but, but how do I find a prior distribution? And so he got up and went to his bookcase and pulled out a book by Cardo Genie, C-O-R-R-A-D-O-G-I-N-I, who you may know if you're an economist.

Um, about the Genie coefficient, which is used in, in, in, especially in the economics. Uh, and Genie turned out to be also a demographer. So I looked at the book and in the back of the book in the last half of the book or so, was amazing data on families throughout the world. A Australia, uh, Africa, the United States on family size and number of boys.

Uh, that was one that is, all of them were boys, was bigger than the corresponding proportion for 11.

Uh, based on that prior probability, of course it went, it went up for the next one. And, uh, uh, mom and I had two girls. Uh, So,

He worked with John Von Neuman as a mathematical statistician. I remember Edwards Lindman in Savage where it was behavioral sciences he worked in, sounds like he was actually an, an applied statistician despite being a genius mathematician.

Uh, the, what the integral equation was that you had to solve in order to do this thing, but it was about a real application, like this thing, um, the distribution of of gender in, in families.

Namely, there's an amount of money that you really want and need to buy an airplane ticket to go home and nothing else matters. And so it's utility, one above that and utility. And then there are some strange things that happen, uh, in the strategy and in particular, bold play is optimal. Um, and so he took this seemingly small problem. And, uh, built finite adaptivity. It made it show why you need finite adaptivity and why it's more important.

Um, and so it was, and, and that was not unusual, where at the end of the road was a revelation about the mathematics. Uh, so, uh, anyway, he was, he was, um, not interested in clinical trials. The one person who was was Frank anco.

Um, and so, uh, you, but you learn over time and you learn, uh, you know, when you've, when you've had enough and you're convinced that you're never gonna win at this, so you go to do something else. Uh, and so I. I said, you know, the two armed version is even more interesting. And so I started working on it, did some dynamic programming, did some computations, and I really became, uh, interested. Um, and so I got Savage interested

And so, I mean, I just, and I knew he was interested in strategy and I knew that this problem would really interest him. I didn't know the, the name Bandit. Eventually I knew that it was called Bandits. It's a very hard problem. Uh, suffice to say that the computer scientists call it NP-hard Uh, suffice to say that, uh, Peter Whittle, who was uh. Uh, uh, worked in World War ii. It was a, a statistician, uh, worked in World War II on, uh, strategic things.

Said that, um, they talked in, in the UK about out making a one sheet description of the bandit problem and dropping it on Germany. So that the mathematicians would get hold of it and waste their time trying to solve this problem.

So I I, in some ways I'm almost your academic brother in this and, um, my daughter Lindsay, um, interestingly her last two choices were CMU and Duke. And if had she gone to CMU, she may have been j Kade student, and so it, it may have been that much more twisted. And interestingly, Lindsay. Was an honorable mention for the LJ Savage Award for her dissertation In Bayesian statistics.

And so it ended up going through five drafts and I tell you that every single one of them, he looked at every single word he recorded what he was doing. You have to understand it's a little bit delicate 'cause you know, of his writing and, and. Uh, eyesight. Uh, but he would record his, what he was reading and explaining to me. He taught me how to write.

After that, it split into two Annals of Statistics, annals of Probability, uh, and it, the associate editor was Tom Ferguson, a very famous decision theorist, and Ferguson didn't send it out for review. He read it himself, he approved it himself. It was published 27 pages and there were no revisions at the journal level. Um, and it was J it was, it was impeccable because of Jimmy who was. Uh, uh, he was a perfectionist. It was his downfall.

Some you mentioned, uh, some other people that might not have been. So, I mean, I loved Savage. I mean, he was like a father figure. He was a father figure for me. Um, and I'd have done anything for him. Um, uh, but other people had been in conversations with him that. Uh, we're not all that positive. I know Herbert s Shernoff, who I think is an amazing person. I, and I love him too. He's still alive. He's 101. Um, and, uh, so I, I said to him he's interested in, in sequential things.

Uh, he was Jake Kaine's advisor,

He and I said, well, but Herman, he, he was human and Chernoff said he had some human characteristics. So it, it, it, and, and, you know, everything comes back to the. The, the eyesight and the way he was treated. Uh, let me just mention, and I said I was gonna say something about Bill Cleveland. Uh, Janice Cleveland was his wife. We had a party one night at, at, at, uh. Uh, Savage's house and, uh, there were the, the, the women, the, the wives mostly, uh, were talking.

Um, and Jimmy came up and, uh, entered the conversation and, uh, uh, Janice Cleveland said, uh, uh, professor Savage we're talking about children and, you know, what are the good ages for children? What's your favorite age for children? And he said, your age. And now people laughed. But it was real from his, it was from, he was serious from his experience. Uh, he didn't want to be exposed to these people that had really done some, uh, nasty things.

I mean, you can imagine the bullying and the like. Uh, so it's a Sergeant Freud. It was a, a very good story. Uh, he died too soon. He died at age 53 from angina.

Anybody have questions about the homework? And, uh, there was a question and the guy asked, uh, really the hardest question in, in the homework. Uh, and it had to do with the acidotic, uh, tail of the normal distribution. And so I showed him how to do it. I started to show him how to do it, and then Savage came in and, um. I went up and I handed the chalk to him. He said, no, you continue.

So I continued and uh, at, I showed him how, how to do the problem, but then I related the problem to some real things and this impressed Savage. Um, uh, and he then asked me, uh, if I would be willing to teach a course. Uh, at Albertas Magnus College, a Albertas Magnus College was in New Haven. Still is. It was, uh, all girls co-ed now. Uh, but they had a statistics course that had not, um,

And he said, Albertas Magnus. The car, the car's name is Big L. And so that's part of the story. But he, and then he kept calling it Big L but then we had a very famous guy, and you've probably heard of him, um, Fred Mueller.

Uh, uh, uh, Jerry Kornfield, I mean, it was a mecca to go to. Um, and so, uh, when he was going back, Savage asked me if I would take him back to the airport, and he wanted to go along with, because they had worked together, he and, and Ello. So the.

And so what you've said now makes it all worthwhile. That it is a useful thing and it, it hearkens back to the encyclopedia that he read. You know, I mean, he was really. A human encyclopedia, but that's too technical a thing. I mean, it isn't, he wasn't technical. He was brilliant and innovative and, uh, uh, uh, I'm, I'm, I'm, I'm glad we had him in the Bayesian world. We probably wouldn't exist, Scott, if it were not for him.