Biostatisticians’ idea has real-world applications

Dr. Lloyd Edwards

Lloyd Edwards, PhD, planted an idea with his R2Beta statistic, and a group of landscape geneticists helped it grow. The results have been harvested and published in the August 2012 issue of the journal Molecular Ecology.

The R2Beta statistic helps researchers assess how important any given predictors are in explaining an outcome. In 2008, Edwards, associate professor of biostatistics at UNC’s Gillings School of Global Public Health, developed the statistic with colleagues Bahjat Qaqish, MD, PhD, associate professor of biostatistics; Keith Muller, PhD, formerly of UNC and now on the faculty at University of Florida; and Russ Wolfinger, PhD, and Oliver Schabenberger, PhD, senior statisticians at SAS Institute. Though other researchers had proposed R2 statistics for a model of analyzing longitudinal data, “ours is the one that’s most comprehensive and easiest to use,” Edwards said.

R2Beta has been used widely in applications in clinical and observational studies; nutrition research; human, plant and landscape genetics; and development of other statistical methods. The statistic allows researchers to assess the strength of association for any particular variable or a whole set of variables.

A group of landscape geneticists led by Maarten J. Van Strien of the WSL Swiss Federal Research Institute used R2Beta in assessing the effect of the landscape on intraspecific genetic structure. Van Strien used R2Beta to choose the best linear mixed model, which is not the use for which Edwards developed it, to examine migration habitat and landscape elements that either inhibit or facilitate gene flow.

“This is a novel use,” Edwards said. “Our statistical method was not aimed at this, but it is being used to develop a new method in modeling landscape genetics. It’s a classic case of translational science.”

Van Strien’s use of R2Beta is a new implementation of the statistic, opening the door for Edwards to create the mathematics to fully support the novel use and giving him a jumping off point for where to take his research next.

“It spurs more development on our side,” Edwards said. “You see the intertwining of biostatistical research with investigative research.”

Van Strien’s article also cited work done by some of Edwards’ former students: Matthew Gurka, PhD, ’04, chair of biostatistics at West Virginia University, and Jean Orelien, DrPH, ’08, founder and chief executive officer of SciMetrika.

Biostatisticians generally act as support for investigators in other fields. In the Van Strien article, however, the statistical method is front and center.

“This article shows we’re not doing just ivory-tower research,” Edwards said. “R2Beta is being used in applications to help answer practical research problems that affect all of us.”