A manuscript describing the EPRV3 Evidence challenge has been submitted and is avaliable at https://arxiv.org/abs/1806.04683 . Suggestions are welcome.
As part of the Third Workshop on Extremely Precise Radial Velocities (http://eprv2017.psu.edu), we formed a breakout group on Computational Methods. The breakout group engaged in coordinated research before, during and after the meeting to conduct a data challenge, where multiple participants/groups compute the Bayesian evidence (i.e., the marginalized likelihood) for models with zero, one, two or three planets for each of ~6 simulated RV data sets. We provided detailed information about priors and noise assumptions, so that everyone could use the same physical and statistical model to estimate the same quantities using multiple methods. Results are described at https://arxiv.org/abs/1806.04683. We hope that the evidence challenge will help the group and the RV community to learn about the relative strengths and weaknesses of various methods, discuss ideas for improving accuracy or efficiency, and form the basis for a short paper summarizing the results and lessons learned.
Sincerely, Eric Ford, EPRV III Computation Methods Breakout Session Chair Ben Nelson, EPRV III Evidence Challenge Coordinator https://groups.google.com/forum/#!forum/eprv3-evidence-challenge https://github.com/EPRV3EvidenceChallenge/Inputs http://eprv2017.psu.edu