Investigate outliers in prediction algorithm

[TomWagg]

Basically just want to investigate what's happening with the leftmost blue objects and rightmost orange objects in the plot above

[TomWagg]

False negatives

• Edge of fields
  • Cases where the real object just about passes from one field to another. Most orbits are only in one or the other
  • 0035981
• Radial velocity sensitivity
  • Location is strongly dependent on RV, large initial range means most orbits are different
  • 000F110 (later nights)
  • Begs the question of how we justify my arbitrary [-100, 100] km/s

[TomWagg]

False positives

• Field Shape
  • Possibly the shape of the field comes into play? (i.e. lost in the corners)
  • 000296A
  • 000D5F1 (night 11)
• Combinations
  • Seems like low probability detections over different nights can combine to an overall high probability. i.e. some orbits have one combination of 3 nights and others have different 3 nights, combine to high probability. I think this is okay?
  • 00000A7
• !???!!?
  • These ones baffle me, surely they should have been detected
  • 0029679
  • 000D5F1 (night 13), maybe magnitudes though?

[TomWagg]

Summary of responses to this after speaking with Mario:

• Account for the camera footprint using RubinSim
• Don't worry about combinations or edges of fields, that's just luck
• Radial velocities are too coarse, we're going to reduce the range