Group Leaders: Jessi Cisewski (Stat, Yale); Eric Ford (Astro, Penn State)
SAMSI Webmaster: David Stenning
Weekly Meeting: Thursdays 12:00-2:00pm ET
Description: This working group aims to improve the statistical methodology for interpreting detections of exoplanets, gravitational waves (GW), as well as using those to infer the underlying population of planetary systems and GW sources. The exoplanets community is particularly interested in developing techniques to robustly detect and characterize planets in the presence of stellar activity from Doppler Surveys for which we do not have a first-principles model. The GW community is interested in detecting gravitational wave sources for which the details of the primary GW signal and/or backgrounds are unknown. Both applications require developing algorithms to efficiently explore high-dimensional parameter spaces and to establish confidence in detections, despite complex and unknown sources of background signals that “noise”. Similarly, modeling either planet formation or the relativistic merger of astrophysical objects is so complex that it is impractical to use a rigorous model for all stages of inference. Therefore, we will develop emulator functions and marginalize over model uncertainties in a principled way. We also anticipate exploring how these methods can improve characterizations of planet masses and densities from transit timing variations measured from photometric time series. Finally, both communities seek to combine information from a survey to characterize the distribution of sources in nature, accounting for measurement biases and incompleteness.
Big Questions:
- How can spectroscopic observations provide information about stellar activity to improve the detection of planets from Doppler Surveys?
- How do we carry out inference when we don’t have complete confidence in our models? How do we marginalize over model uncertainty? In the case of Gravitational Waves (GWs), this applies in two different contexts: inference on individual signals, to extract source parameters such as masses from GWs signatures; and inference on populations, to extract evolutionary parameters such as common-envelope efficiency from multiple observations of a population of objects. For exoplanets, this could refer to either the exoplanet population, or characterizing the masses and orbits of planets from Doppler observations in the presence of stellar activity.
- How do we infer an underlying population distribution in multiple dimensions from a selection-effect limited sample with imperfectly modeled selection effects and measurement uncertainties? In the context of exoplanets, the most obvious example is “What is the occurrence rate of various types of planets and planetary systems?” In the context of GWs, this applies to modeling subpopulations of different flavors of compact binaries (neutron stars vs. black holes, isolated binaries vs. dynamically formed systems) based on a set of observations of individual binaries, each with imperfectly modeled parameters.
- If we have very expensive simulations which we can only afford to run a few times, how do we decide in a pseudo-online regime (as exploring the parameter space) where to run them to optimize overall inference accuracy? For GWs, this applies to both inference scenarios above: numerical relativity simulations to produce accurate GW models in one case, Monte Carlo population synthesis simulations in the other. For exoplanets, this could apply to either the exoplanet population or simplified planet formation simulations that are being used in a population synthesis sense.
News and Updates: Coming Soon…
SAMSI Directorate Liaison: Sujit Ghosh
Questions: [email protected]