Why
do you think your paper is highly cited?
Recently there has been much discussion in phylogenetic
and population genetics research about which methods are
best for estimating population genetic parameters. My paper
compares the performance of a maximum likelihood (ML) and a
Bayesian estimator for such parameters.
Does
it describe a new discovery, methodology, or synthesis of
knowledge?
|
“The
paper highlights a pragmatic non-philosophical
comparison between two methods (maximum
likelihood inference and Bayesian inference)
commonly used to estimate population genetic
parameters.” |
|
|
The paper highlights a pragmatic non-philosophical
comparison between two methods (ML inference and Bayesian
inference) commonly used to estimate population genetic
parameters. It elaborates on some of the problems of each of
the methods in the context of population genetics
applications that use Markov chain Monte Carlo methods to
estimate quantities of interest.
Would
you summarize the significance of your paper in layman’s terms?
Comparisons of ML and Bayesian estimators are most often
made using different programs. I describe the program
MIGRATE, which can be run in both estimation modes. This
allows a more direct and fair comparison of the methods
because most of the computational machinery is the same in
both methods.
ML and Bayesian methods both have their strengths, but
for an exploratory use of such programs one can probably
understand problems with assumptions and problems in the
data more easily with the Bayesian approach. For the
experienced user the difference between the methods seems to
be more a matter of philosophy.
How
did you become involved in this research, and were there any
particular problems encountered along the way?
I started out as an evolutionary biologist who was not
satisfied with the then-current methods for estimating
genetic migration rates. Work with Joseph Felsenstein and
Mary Kuhner at the University of Washington allowed me to
develop a better migration rate estimator. In my position in
the School of Computational Science and the Department of
Biological Science at Florida State University, I find an
ideal mix of users of such programs (biologists) and
developers (computer scientists, mathematicians) that allows
me to further develop both new methods and extensions of old
ones. I hope that these are useful for other researchers.
Are
there any social or political implications for your research?
I believe there are no large-scale social and political
implications of my research, but my and other researchers’
programs are increasingly used to infer past population
sizes of endangered species. Some of these new estimates
differ markedly from estimates that are currently believed
to be correct, which has led to interesting reactions.
Peter Beerli
Computational Evolutionary Biology Group
School of Computational Science (SCS)
and Biological Sciences Department
Florida State University
Tallahassee, Florida, USA
