12/17/21: Regular meeting canceled; notes on another meeting instead

DB met in the Boston area with our contact to discuss the data

Here are some notes from the meeting.

• A way to deal mathematically with things that have no upper bound, like lifetimes of spacecraft that are still operating, is the Kaplan-Meier bound or Kaplan-Meier approximation.
• The problem of overfitting might be addressed in part using PCA. 
• A problem with the half life approach is that a satellite might remain indefinitely in a non-failed status simply because definite information about failure never arrives. We should feel free to ask about specific cases, but this makes sense more for high value data points like for the interplanetary exploration project than for Earth satellites.
• What is a flyby? One way to define this a "Hill sphere" which is the approximately spherical shell around a body which is the boundary between where its gravity predominates and where the gravity of another body predominates. Lagrangian points would be on this sphere, if I understand correctly.
• When time between launches gets too high, personnel turnover prevents experience in one launch from helping the next launch, so failure rates will tend to increase.
• Satellite S25744 is one where the end date is very vague, "2010s?" How does VJB handle that one?
• Last known radio contact with satellites is a key data point in determining lifetimes. Tilley is someone who has valuable primary data of this time. 
• Date formats are often of necessity vague. See planet4589.org/space/gcat/web/intro/vague.html for one way to do it.
• Why not do automatic outlier detection of individual fields and of relations among fields as a way to identify errors? Maybe a student could be found who is interested in doing this. Kratschmann is interested in error detection on this data set and might be worth tracking down if such a student is found. Maybe he already does it that way?
• Lifetimes are often affected by more or less imponderable or highly domain specific considerations. For example TRW makes good spacecraft, while some other manufacturers make spacecraft that are less durable. There is a maker in Russia that has made a lot of craft of lower quality. Satellites launched by universities are less reliable, potentially, because some of them are designed by students (as a conjecture).