How Did I Get Here?
which led me to empirical Bayes and, eventually, research within
So, that is the story of what I do. But, how did I get here? I fol-
the Bayesian paradigm. Finally, Adrian Smith was most influen-
lowed a fairly typical course for my time, starting as an under-
tial to me. Following conversations with him in the mid- to late
graduate major in mathematics. In my senior year, I took an
’80s, I arranged a sabbatical to work with him at Nottingham
introduction to mathematical statistics course from the book by
University. I arrived seeking to use his numerical integration
the same name written by Robert Hogg and Allen Craig (embar-
software for certain empirical Bayes problems and left with the
rassingly, using the first edition, spanning all of 240 pages). The
Gibbs sampler!
material was so elegant, I was smitten, and I applied to graduate
In this regard, I was fortunate to be involved in a seminal contri-
programs in statistics, moving to Stanford—as far from New
bution to our field. Indeed, working with Smith, it was a remark-
York City as I could go.
able feeling to come upon the Gibbs sampler from the 1984 paper
I thoroughly enjoyed my four years at Stanford. (The late
of Stuart Geman and Donald Geman, “Stochastic Relaxation,
1960s was a remarkable time for a statistics department in the
Gibbs Distributions, and the Bayesian Restoration of Images,” and
San Francisco Bay area—and for the country.) I emerged with
recognize it was better suited for handling Bayesian computation
training as a mathematical statistician, really the only path avail-
than it was for its original purpose—sampling Markov random
able at that time. As I have told many people, I had a ‘wasted
fields. In fact, for the first half of the 1990s, I spent almost all
youth’ in some ways, not discovering until the late 1980s that I
my time writing as much about the use of Gibbs sampling and
was born to be a Bayesian (and, with it, to be a modeler).
Markov chain Monte Carlo algorithms for Bayesian model fitting
People who influenced me include my adviser from Stanford,
and model determination as I could. These were heady times for
Herbert Solomon, a wonderful, generous man who was particu-
those of us involved, as we realized we were elaborating on a tech-
larly good at recognizing how people could best contribute to
nology to successfully address the problem of Bayesian compu-
projects. He was one of the earliest ‘team builders’ in our field.
tation—to replace high-dimensional integration with simulation
Charles Stein, who was also from Stanford, influenced me in
from high-dimensional distributions.
a distant way. I believe I hold the record for the most courses
In the mid 1990s, I started moving from research confined to
ever taken from Stein by a single student. His beautiful, deep
computational matters to the analysis of spatial and spatial-temporal
thoughts in decision theory led me down that research path,
data. Again, I was fortunate. Spatial statistics had struggled as
22 AMSTAT NEWS SEPTEMBER 2008
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84