We had the big battle about the form for a likelihood for exoplanet search and characterization late in the day, with Baines (Davis) leading the discussion. There was a huge disagreement about the realism of the model—as there always is—with some saying we should split stellar and spacecraft contributions to the lightcurve variability, and some wanting to mush them together. The former is the Right Thing To Do (tm) when you are going hierarchical, because it permits you to pool data from multiple stars on the same part of the detector (for the instrument model) and multiple stars of the same type scattered around in time and space (for the stellar model). That said, the mush-together option might be the right thing to do for an effective model where you want or need to treat every lightcurve separately. We chose the mush, and Baines and the gathered worked out some ideas about what kinds of models for stellar variability and spacecraft artifacts might work.
In the rest of the day, the team worked on selection of target stars, injection of false signals, running standard filter-based and fitting-based de-trenders and co-trenders, investigation of wavelet transforms, and statistical properties of stellar variations.