Stats in the Family Make a Difference in the World
What does research in the areas of endangered species and evolution have in common? Professor Charles Geyer of the University of Minnesota can tell you: statistical analysis. Throughout his career as a professor and also a researcher, Geyer has created and developed specific models used for statistical analysis to find meaning in data.
As an undergraduate, Geyer majored in physics but found himself uninterested in attending graduate school. Upon deciding not to go further in the field, Geyer became interested in statistics, often dabbling in it partly due to his family’s influence. Geyer’s sister, Ruth Shaw, is a professor of evolutionary genetics at the University of Minnesota, and her research exposed him to a new area of statistics: life history analysis. While Geyer participated in some of her projects, his father was also a chemist whose work interested him. “Stats is a family project,” Geyer says.
At the age of forty, Geyer decided to go to graduate school due to his developed interest in statistical research. Attending the University of Washington, he knew his experience with his sister’s projects had prepared him well. While there, he became involved with seemingly unrelated research projects; however, as he began to connect the dots, he saw how similar models could be used to solve unrelated problems.
The California Condor and Markov Chain Monte Carlo
During graduate school, Geyer became involved with a research project at the San Diego Zoo involving the California condor, North America’s largest bird. The species, however, is endangered and the zoo’s research director needed help looking at the genetics of the captive breeding population. At the time, the entire species was part of a captive breeding program at the zoo, but has since been released. The survival of this impressive bird was dependent on understanding this essential part of their behavior and analyzing the DNA fingerprinting data, a new tool at the time.
From the analysis, Geyer and team saw how location affected the bird’s breeding structure. By using statistical models and approaches to the project, they discovered that the population structure was closely tied to geography; the birds chose not to fly as far as they could.
In the conclusion he found, Geyer happened to come across the Markov Chain Monte Carlo, a very specific area within spatial statistics. This area become extremely popular in the 1990s and was universal in simulating any population distribution model. “It was total luck that I stumbled on it,” Geyer notes. “I became an expert on something not because I planned to but because it just happened.” His research eventually turned into a paper still widely cited today.
Darwinian Fitness and the Aster Model
Geyer, Shaw, and their coworkers have also participated in research involving an area called life history analysis. This area, which involves studying evolution and the Darwinian fitness of organisms, was in a problematic state before the development of the aster model. “Biologists knew what data they needed to address the evolutionary and ecological questions of scientific interest, but there did not seem to be any suitable statistical methods to analyze such data,” Charles says.
The idea of aster models was developed around 1980 by Geyer when Shaw was a graduate student. However, it wasn’t until 2005, when both Shaw and Geyer were professors, that they applied the aster model to a project. This project, known as the Echinacea project, collected life history data on the native prairie wildflower Echinacea angustifolia. “They really needed to do valid statistical analyses,” Geyer notes.
Unable to find the original idea from 1980, Geyer reconstructed the model to make it more generalized. With the aster model, the Echinacea project was able to make sense of its life history data. On top of this, he provided software to complete aster analyses in seconds. This model has resulted in more than 20 research papers from their group and over 100 other papers that cite them. “Aster models are on their way to becoming a widely used scientific tool,” Geyer says.
Although the aster models were proven effective in making sense of large data, Shaw and Geyer are far from done with them. Both professors are on sabbatical this year with that time devoted to further development of the aster model.