Advanced Statistics in the Search for Planets

The search for planets outside our own Solar System (exoplanets) has been very successful. One search method looks for changes in brightness as a planet crosses in front of its parent star. Another looks for small wobbles in the star's motion. Both methods are hurt by noise that can overwhelm the weak planetary signals. A new approach uses advanced statistics to clean out the stellar signal and find planets in the residual signals. This could transform the field by uncovering interesting planetary systems. It introduces modern methods into astronomy. This will improve calculations of planet populations. It will also benefit all studies that concern the time domain, especially those with complex variability. This award helps support training science teachers in the exciting topics and methods of astronomy.

The search for exoplanets with photometric transit and spectroscopic radial velocity surveys has been very successful to date, but progress is stymied by the presence of signals from stellar magnetic activity or other nuisance noise sources that can overwhelm the weak planetary signals. A new approach has been developed that uses parametric statistical techniques to remove most of the stellar signal, and then searches for planetary signals from the residuals. Preliminary work on the Kepler mission database is showing excellent results, and this study will extend these methods to ground-based surveys. It is clear that astronomy has not taken sufficient advantage of modern advances in statistical methodology, such as the autoregressive modeling of time series developed in signal processing and econometrics. This project will start by investigating how well these sorts of models work on typical ground-based cadences, and then test the successes on a photometric database of several million stars. Autoregressive models will also be tested on a densely cadenced radial velocity survey of bright stars, chosen for the confusing effect of magnetic activity on correlated jitter. All of the methods developed will be widely promulgated, and can serve a wide range of other time series analysis problems in astronomy. The exoplanetary community will benefit from the discovery of new candidate planets, uncovering interesting planetary systems, and improving calculations of planetary populations. The time domain astronomical community will benefit from using autoregressive methods to treat the complex variability behaviors of cosmic populations.