Apply single IFO cut only when more than 2 IFOs are used by PyGRB

[pannarale]

The current version of PyGRB requires the single IFO SNR to exceed a threshold (provided by the user) in at least 2 IFOs, whenever the analysis uses more than 1 IFO.

CohPTF used to apply the cut on the two most sensitive detectors when more than 2 IFOs were involved in the analysis.

This change restores the "more than 2 IFOs" requirement. It does not do anything about the "most sensitive" concept because:

1. the ranking of sensitivity could be carried out in more than one way (at each sky point of the search, at the centre of the grid, averaging over the grid) and this would require a systematic study;
2. perhaps more importantly, more sensitive does not imply anything about glitchiness.

This change affects: PyGRB.

This change is being open to discuss whether to bring the current code closer to the original approach. There is no systematic comparison before and after this change on N GRBs with 2 IFOs. I am also pretty sure that there was also nothing of this sort that motivated applying the cut whenever more than 1 IFO was available.

• The author of this pull request confirms they will adhere to the code of conduct

[titodalcanton]

This brings back memories! I would love to do a proper study on it. WIthout that, I have no particular opinion, just two questions:

• How did cohPTF define "most sensitive", and would it be easy to reimplement the same?
• Why not require a single detector to exceed the threshold? That seems the simplest thing to try, especially in light of how many double-but-effectively-single events we have seen so far.

[pannarale]

Looking at this line, the help message is in line with your second item, but I distinctly remember going through lines of code with Andrew where the whole "2 most sensitive" thing was coded up (your item 1). Let me dig through some code and get back to you.

Edit 1:

• The filtering stage would do this
• The post processing would calculate at F+ and Fx at the RA and DEC of the trigger, rank the IFOs on the basis of the antenna response, and demand that the threshold is exceeded in the two with the highest responses. This is the kind of thing we would be after ultimately.

Edit 2: I should add that the single IFO SNR cut on the 2 most sensitive IFOs was originally formulated for equally sensitive IFOs, i.e., with the same PSD.