Feb 9, '22

Action Items from the last meeting:

• Followup on

  • TOI 178 (multiplanet)
  • TOI 237 (potentially multi-modal fit?)

• Look at phase-plots and divide erroneous fits into categories (eg. wrong period init https://github.com/dfm/tess-atlas/issues/101)

• Data:

  • use the quality flag --> hardest [done]
  • remove EXOFOP source=SPOC filter [done]

• Rerun analysis with above changes to data (and other changes that have been made)

Case Studies

• 178 -- GP overfitting!
• 237 -- multimodal posterior -- what can we do?

New items:

• Single transit analysis is broken: https://github.com/dfm/tess-atlas/issues/29
• GP overfitting: https://github.com/dfm/tess-atlas/issues/183
• Can we compute a statistical value to determine how well we've fit the data?
  • Maybe how gaussian the residuals are?

Action items

• fix the pages! Hard to share output atm
• Diagnostic plotting:
  • plot the residual plot after the initialization
  • plot raw data - transit and overlay GP
  • don't use median value -- that is uninformative -- maybe plot all the LCs with lower alpha?
• plot the binned data https://gallery.exoplanet.codes/tutorials/lc-gp-transit/
• Debug the GP weirdness for 103 -- is the GP fitting data- transit model? Whats going on?
• Next week: look at model with dan to help with single transit systems