Fermi Gamma-ray Space Telescope

Pass 8 vs. Pass 7: How Statistically (In)Dependent Are They? (And Why Should You Care?)

L. Baldini
M. D. Wood, on behalf of the Fermi-LAT Collaboration

Abstract:

Pass 8, publicly released in June 2015, improves over the previous iterations of the Fermi-LAT event-level analysis in all the metrics relevant for high-level science analysis. Particularly, compared to the 'equivalent' Pass 7-reprocessed event class, the Pass 8 SOURCE selection provides a 25--30% acceptance increase between 1 and 100 GeV. This is not equivalent to, for any given integration time, Pass 8 adding 25--30% more events to the corresponding Pass 7-reprocessed data set. Since the Pass 8 event reconstruction is fundamentally different, the basic characteristics of individual events---including direction, energy and photon-likeness---can change, causing photon candidates to migrate in and out of ROIs and/or energy bins and across the threshold for any specific photon selection (the latter being true particularly for misclassified cosmic rays). One interesting consequence is that the fraction of shared events between 'equivalent' event classes in Pass 8 and Pass 7 can be significantly lower than one might naively expect based on the ratio of the relative acceptances. We illustrate the concept using the LAT combined dwarf analysis, where the 6-year Pass 8 constraints at 100 GeV for the representative b-bbar channel are 5 times lower than for 4 years of reprocessed Pass 7 data.