28
April
2009
|
18:00 PM
America/New_York

Scientist Wins Top Statistics Prize

By Pam Frost Gorder

For finding the common ground among many different fields of science, an Ohio State University researcher has won the highest honor in his own field of statistics.

The Committee of Presidents of Statistical Societies (COPSS) has awarded Noel Cressie the R.A. Fisher Lectureship for his “meritorious achievement and scholarship in statistical science and highly significant impact of statistical methods on scientific investigations.”

Noel Cressie

Cressie, Distinguished Professor of Mathematical and Physical Sciences, is the founder and director of Ohio State’s Program in Spatial Statistics and Environmental Statistics.

Over the last 20 years, he has developed new statistical methods for studying everything from the growth of the ozone hole, to the dynamics of flu epidemics in Scotland, to the flow of Greenland’s glaciers.

While these phenomena may all seem unrelated, Cressie has shown that spatial statistics can be used to study all three, as well as many other areas of science, including physics, ecology, epidemiology, agriculture, and neuroscience.

He will present his Fisher Lecture at the Joint Statistical Meetings in Washington DC in August this year.

Jeremy Taylor, chair of the COPSS award committee, explained why Cressie was chosen for this honor.

“His work has truly revolutionized spatial statistics, not only in terms of advances in theory and applications, but in the sense that spatial statistics is now considered to be a fundamental part of the statistics curriculum,” Taylor said.

Spatial statistics has existed in one form or another since the first explorers began to map the world. Their notes on the plants, animals, and even weather they encountered along the way are all early forms of spatial data. But spatial statistics didn’t take shape as a formal discipline until after 1991, when Cressie published his book Statistics for Spatial Data.


Cressie quantifies uncertainty about how data vary in space and time, and combines it with the results from rigorous, controlled scientific experiments when available.  This hybrid technique gives scientists “wiggle room” when they want to study complex systems where some data are missing.


With that book, he showed that many diverse methods for analyzing spatial data could be combined into a single toolkit for scientific research.

Doug Wolfe, professor and chair of the Department of Statistics at Ohio State, said that Cressie’s expertise in this area puts him in constant demand to serve on advisory panels for agencies including the Department of Energy, National Research Council, National Science Foundation, and Environmental Protection Agency.

Wolfe called the Fisher Lectureship “a singular and very much-deserved honor that rightfully recognizes Noel as one of the contemporary giants in our field of statistics.”

COPSS represents three North American statistical societies -- the American Statistical Association (ASA), the two North American Regions of the International Biometric Society, and the Statistical Society of Canada -- and an international society, the Institute of Mathematical Statistics (IMS). Cressie is a Fellow of both the ASA and the IMS.

The committee awards the Fisher Lectureship in honor of Sir Ronald Aylmer Fisher, whose analyses of agricultural and biological data during the first half of the 20th Century laid the foundation for much of modern statistics.  In his work, Cressie uses statistical methods that meld Fisher’s techniques with those of Thomas Bayes, an 18th Century mathematician whose work is currently enjoying a renaissance.

While Fisher maintained that controlled scientific experiments were the only legitimate source of statistical knowledge, Bayes looked to other, uncontrolled sources and a scientist’s knowledge about a subject from past experience.  All sources are more or less uncertain, and it is the quantification of these uncertainties that are the building blocks of a Bayesian analysis.

Cressie’s work merges both ideas. He quantifies uncertainty about how data vary in space and time, and combines it with the results from rigorous, controlled scientific experiments when available.  This hybrid technique gives scientists “wiggle room” when they want to study complex systems where some data are missing.

Today, technology is enabling scientists to gather more data than Fisher could ever have imagined. For instance, when studying atmospheric particles called aerosols -- a principal driver of global climate -- scientists can obtain massive amounts of data from satellites. But satellite images often contain blank spots where no data were recorded.

With Cressie’s methods, scientists can combine current and past data with what they know about how the data vary in space and time to fill in the blank spots. They define a statistical measure of a map’s uncertainty and ultimately produce a global map that minimizes this uncertainty measure. The final product is both the map and its uncertainty measure.

With his Fisher Lectureship, Cressie joins the ranks of John Tukey, the father of modern data analysis; C.R. Rao, a giant of theoretical statistics and formerly a visiting professor at Ohio State; and Sir David Cox and Bradley Efron, both leading innovators of statistical methodology and its applications in the sciences.

#

Contact: Noel Cressie, (614) 292-5194; Cressie.1@osu.edu

Written by Pam Frost Gorder, (614) 292-9475; Gorder.1@osu.edu