Signal efficiency calculation

Hi Saumyen,

> In the cut-flow table for 'Initial (no cut)' case, Signal (S) and Background (B) are respectively, 1.04 and 2.07. Do I understand it correctly that these are basically the number of signal and background events before any cuts are implemented?
Yes they are.

> I noticed that this is the same as 'Events kept: K' when due to a particular cut 'Rejected events: R' is 0.0.
If you reject nothing, then the numbers do not vary. Would you expect a different behaviour?

> Now for this case in the input datasets, the event numbers for Signal and Backgrounds are respectively, 36304 and 42768. Then my question is why these values of 'S' and 'B' are coming so small? So, basically I wanted to know how these 'K' and 'R' or 'S' and 'B' are calculated in MA5?
It depends how the S/B ratio has been calculated. There are several formulas to calculate this number, the default being S / \sqrt{S+B}. You can change them via the "fom formula" option.

Best regards,

Benjamin

Hi Benjamin,

Thanks a lot for your reply and I'm extremely sorry for the so very late response. I was on vacation. :P

>If you reject nothing, then the numbers do not vary. Would you expect a different behaviour?

No no. This I said to tell you how I inferred what I said. Now, my concern is that even if there are a large number of events in my dataset (almost 37K) why I'm getting such small values of 'S' and 'B' like 1.04 and 2.07, respectively when there is NO CUT applied.

>It depends how the S/B ratio has been calculated. There are several formulas to calculate this number, the default being S / \sqrt{S+B}. You can change them via the "fom formula" option.

This is also fine. I had chosen "fom formula=4", i.e, S / \sqrt{S+B}. Right? But my concern is the same that how these 'S' and 'B' and/or 'K' and 'R' are being calculated that it's coming so small even if I have a large number of events.

Thanks and regards,
Saumyen

Hi Saumyen,

No worries for the delayed answer :)

> This is also fine. I had chosen "fom formula=4", i.e, S / \sqrt{S+B}.
> Right? But my concern is the same that how these 'S' and 'B' and/or 'K'
> and 'R' are being calculated that it's coming so small even if I have a
> large number of events.

Having a large number of events is not sufficient. The S/\sqrt{S+B} ratio depends on
(a) the cross section of the signal and background processes (a reweighting is applied)
(b) the considered luminosity (the larger it is, the better it is to get a large significance).

Cheers,

Benjamin