CHICOS Analysiscut2

Running analysiscut2 on old data:

Current state of autoFit4:

• Cut on radius at 8 km.
• Cut on time at 25 us.
• After first reconstruction, deletes hits outside 2-sigma time errors. (There is no requirement that the reconstruction match the intensity well.) Repeats reconstruction and checks again. (Generally one iteration is all that is required.)
• Forces 2-sigma errors to be at least 2us for the purpose of selecting hits.
  *NEW* Uses time spread based on TDF.
• Ignores showers that have fewer than 3 match sites that reconstruct within 2-sigma error bars. (Currently does not check whether the trigger site reconstructs well.)
• Requires 2 sites be fit within 2-sigma intensity error bars, with intensity > 0.1 (so fitting 0 to 0 doesn't count).
• Rejects showers if they are only Chiquita showers after bad sites have been removed.

The new code accepts hits only if they are within a timespread calculated from the TDF. The timespread is the time at which ~95% of the particles have passed. At the moment this calculation starts at the plane of the shower and so is larger than it would be if you started at the shower front (t=a).

To get an idea of how this code performs, I ran it on the current set of 120 Best Candidate showers. Only 76 pass now. Below is the list of showers which USED to pass and now fail:

• 03-03-03
• 03-01-04
• 12-12-04
• 02-01-05
• 03-11-05
• 06-28-05
• 07-05-05
• 09-20-05
• 11-08-05
• 12-15-05
• 12-26-05 One of the two on this day. (Didn't check.)
• 12-26-05b
• 01-02-06
• 02-23-06
• 02-26-06
• 03-08-06
• 03-10-06
• 04-16-06
• 06-17-06
• 09-12-06
• 10-26-06
• 02-03-07
• 02-27-07
• 03-05-07 One of the two on this day. (Didn't check.)
• 03-05-07b

I think it is taking out too many ok showers. Quite likely this is because the timing error is not just the intrinsic width of the shower but also the uncertainty in core location and other parameters. Taking the minuit errors for shower parameters and turning it into a timing error seems unneccessarily complicated - should probably just go back to using the minuit-generated errors, since they ought to take all that into account.