Improve Duplicate Removal Running Twice and Add to Plot

[dally96]

Make some minor adjustments to the duplicate removal code where we fill the truth map twice:

• Exclude tracks that are already true in the truth map.
• Move bin edges over by half of the bin size.
• Should have n-1 bins when moving bin edges.

[dally96]

[trholmes]

OK so it looks like your data still shows that around 70% of the tracks that are input to the DR are in fact duplicates (if this many are being removed). Can you confirm this by looking at the duplicate fraction for the output in one of the samples where you turned off duplicate removal?

[dally96]

For when DR is turned off, the duplicate fraction for output tracks is 89%.

[trholmes]

Hmm.. I don't quite understand why those are such different numbers (although indeed that is very high!). Can you think about how to demonstrate that all these numbers are consistent?

[dally96]

You mean run DR with a large number of bins, and see if the duplicate fraction is around 89%?

[trholmes]

No I mean this: Your fraction removed for small numbers of bins is 71%, while the overall duplicate fraction is 89%. Given the number of duplicates remaining at the end, do those two numbers make sense?

[dally96]

Then, I would have to check that the fraction of number of tracks removed + the number of output tracks marked as duplicate would equal 89%.

[dally96]

With just counting NMatch - 1, the duplicate fraction is 65%, but if I do what I suggested above, when we enable DR, the duplicate fraction is between 71% and 72%.