Sir David Cox has died
rss.org.uk
rss.org.uk
Edit: I have to edit my comment because calling him an absolute legend doesn't feel like it was doing him justice. I'm not kidding, I don't think I can state how important he was. He's in the top 5 statisticians of all time- everyone's ranking of top 5 varies but Cox is in the top 5 on almost every statistician's list- and will probably still be top 5 until the future is perfectly predictable and there is no need for probability and statistics anymore. There may never be someone that is as productive and with as impactful discoveries. Sure, there will be people that write more papers, but nobody will write as many great papers. Sure, there are other great statisticians that have a few great discoveries named after them, but there are none that have as many great discoveries named after them.
If you consider the first 25 as useless/building stage of life and the next 35 as contributing/productive part. This man lived another 38 years after that (an entire new useful life span for an average human).
I never thought I’d read a comment where someone just called their youth worthless. Aside from being reductionist and utterly devoid of empathy, there’s an entire industry built around life extension that disagrees with you.
For someone who lives/interacts with David Cox everyday, the loss is immense.
But OP was referring to 'loss to the society'. When you bring in 'loss to the society', then it is inevitable that you measure their net-positive-contributions-to-society.
Unless you are an outlier (<0.0001%), 25 and below is usually where you are sucking away resources without contributing back (easily measured by net worth or artifacts or offsprings). 25 - 60 is when you are peak contribution (offsprings, net worth, artifacts), after 60 most stop producing (offsprings, artifacts, wealth accumulation). Yes there are outliers to these too.
Now to the controversial. Yes, offspring creation is (the ultimate) productive service to the society. It's the *only* way to keep humanity not going extinct.
Thought experiment: Tomorrow if everyone decides to stop offsprings, humanity is guaranteed to be extinct in 100 years. However, if everyone decides to have a baby or two, humanity will chug along for at least quite a few hundred years (Climate Change is not going to make us extinct)
I read his principles of statistics books and I would recommend it to anyone doing data analysis.
The information age has advanced so rapidly, we hardly realize that almost everything that had been invented, was by people just passed or still alive. When I was in college, Claude Shannon still attended information theory seminars. He invented all of digital communications (or nearly so).
For the younger generation, a logistic regression is a neural network without a hidden layer :)
- Theory of Design of Experiments is an excellent and non-technical book on experimentation and data collection. Still important in an era of big data!
- Analysis of Survival Data is the book on the Cox model. It is quite technical. I own this book for historical reasons, and I'm glad I read it, but I now recommend Therneau for serious survival statistics.
- Theoretical Statistics w/ Hinkley is still a great mathematical statistics book. Excellent math-stats books are uncommon now, and Cox was a master.
Cox also wrote a few books on stochastic processes that are well regarded, but I'm unfamiliar.
I most recently read Celebrating Statistics: Papers in honour of Sir David Cox on his 80th birthday which is a very thoughtful collection.
From Cosma Shalizi's review [0]:
> D. R. Cox published his first major book, Planning of Experiments, in 1958; he has been making major contributions to the theory and practice of statistics for as long as most current statisticians have been alive. He is now in a reflective phase of his career, and this book, coauthored with the distinguished biostatistician Christl A. Donnelly, is a valuable distillation of his experience of applied work. It stands as a summary of an entire tradition of using statistics to address scientific problems.
The story about that goes back to the time when I was still in England, and David Cox and I were both on the research committee of the Royal Statistical Society. Various people remarked on the fact that Box and Cox were both on the committee and said that we should write a paper together. My recollection is that either David said to me or I said to David, "You know, perhaps we ought to take them up on that." And so, "Yes, OK, let's do that. And what should it be about?" Well, we both knew the story about Box and Cox. This is a story about Box living in a room during the day and Cox living in the same room during the night, and neither of them knew that the landlady was giving the room to two people and getting two rents rather than one. So we said, "Well, obviously the thing to write about is transformations." And that was all we had to begin with. [Laughs] We did something about it. We got as far as the part about the Jacobian, and if you've read that paper ["An analysis of transformations," J. Roy. Statist. Soc. Ser. B 26 (1964) 211-252] you'll know that this part is a little bit tricky, even controversial. We didn't know quite how to deal with that bit, and so it got put to one side. The time passed, years passed; I went to Princeton, and then to Madison. After I had been at Madison a short while, the way I remember it is that David wrote me a letter and said, "Hey, you know that thing we were doing? I think if we went this way, we would be able to do it." And I wrote back and said, "Yeah, I think so," and what about this and what about that, and so on. And so we started working together again on it, and finally we got the paper out.
DeGroot: A classic paper. So it really did start with just the authors' names and built up from there.
Box: Yes. I think we also had sort of a slight conspiracy because I think we both felt it was a bit funny about the Box and Cox notion. David was keener on the likelihood approach and I was keener on the Bayesian approach. So we put both in, and when the discussion came up various people said, "Perhaps the authors would like to say whether they both agree on this" and we were careful not to say. So there was sort of a slight Box-Coxish uncertainty about the whole thing which we thought was amusing. -------
One of the most cited papers in Statistics, the Box-Cox transform (http://www.econ.uiuc.edu/~econ508/Papers/boxcox64.pdf , 18000+ citations) is a power transform just about everybody has employed at one point or another when dealing with non-Gaussian data (builtin function MASS::boxcox in R)