25 July 2025 Friday

W: 6 DW: 2 OT: 1 ex

• See logs

• See above

• 3s and 4s … but only half court …

• See above

• Looking into ZZ and ZH
• May have to revert to previous SvB

Hi Guys,

Alejandro and I have been thinking about Vivek's concern about the scaling of the sensitivities between Run2 and Run3. If I understand correctly, one thing that makes the comparison harder for the boosted analysis is that there is a significant statistical uncertainty in the central value of the 95% CL. In both the resolved and boosted analyses there are large statistical uncertainties on the expected limit (the Brazilian bands) however in the resolved case, these are almost entirely due to the expected statistical uncertainty on the 4b SR data; the statistical uncertainty on the background is negligible. In this case, it makes sense to directly compare the values of the central values, ignoring the large Brazilian-band uncertainties.

However in the boosted case, I guess you have a significant statistical uncertainty on the expected background coming from the TF fit in category 1. This would lead to a significant statistical uncertainty in the central value of the 95% CL. Here, when you compare expected limits between the two analyses, I think you want to take this uncertainty into account. Im not sure if there is an easy way to get the size of this uncertainty directly from combine. (One way to hack it would be to do you blinded-asimov fit, but set the statistical uncertainties on the asimov data to be very small).

Once you have this uncertainty, you can then answer Vivek directly by checking if the scaling seen in the boosted expected limit is statistically-significantly different from that seen in the resolved.

Does this make sense?