Extrasolar planet searches have revealed an unexpected diversity of planetary systems and provided insights into the mechanisms of planet formation. Recent transit timing observations are very powerful for detecting Earth-mass planets around stars with transiting giant planets. However, relatively little attention has been paid to the statistical methods that will be necessary to make robust planet detections and accurate measurements of planets' masses and orbital properties. In this project, Dr. Ford and collaborators will work to provide the needed statistical foundation for transit timing observations to detect and characterize Earth-mass planets using existing ground-based observatories. Statistical and computational methods are particularly important because the transit timing signature of a low-mass planet is typically dominated by the mutual gravitational interactions between the low-mass and giant planet, rather than by the direct effect of the planets on the star. This requires computationally expensive N-body integrations of a highly non-linear system, which makes computationally efficient statistical methods essential. This program will research algorithms for rapidly exploring high dimensional parameter spaces with an adaptive surrogate Bayesian model, and for efficiently calculating Bayes factors from posterior samples generated via Markov chain Monte Carlo. It can be expected to lead to robust detections of low-mass planets and constraints on planet formation theory.
This research will contribute general statistical algorithms to the public domain. They will be applicable to other types of current and future extrasolar planet searches as well as a broad range of problems in time series analysis and Bayesian model comparison. Through this project, a postdoctoral research associate and graduate and undergraduate students will develop extensive expertise in both statistics and astrophysics by participating in cutting edge research. An astrostatistics visitor program, including experts from underrepresented groups, will train a broader array of students in modern statistical methods for the physical sciences and will stimulate interaction and collaboration of astronomers and physicists with statisticians.
