search

XENON1T / pax

The XENON1T raw data processor [deprecated]
BSD 3-Clause "New" or "Revised" License
16 stars 15 forks source link

Change S1 Decay Time in Sim #665

Closed zhut19 closed 6 years ago

zhut19 commented 6 years ago

Change S1 decay time to match data Step compare data with fax

width_compare_1 08

Maximize Log Likelihood

maximum_ll
feigaodm commented 6 years ago

Thanks @zhut19, Looks solid! The mean value should be directly useable. We'd like to evaluate the undertainty of this parameter as well, from you likelihood curve. I think your likelihood value could be wrong. Can you write down you likelihood? The easist would be to assume gaussian error bars on sim and data, and calculate chi^2/ll values.

zhut19 commented 6 years ago

@feigaodm I took data means as Y0, fax means as Y log likelihood = sum( [-(Y-Y0)**2] ) (As I don't know the uncertainties on the means)

feigaodm commented 6 years ago

@zhut19 Yeah, I think you derived the quantiles from simulated and real data. You can calculate the uncertainty on the quantiles, I don't remember a formular to do this, at least you can use bootstrap technique. (Or you can google a way to calculate it, I did it but couldn't find a easy solution). Then you can use the uncertainty to recalculate your likelihood.

pdeperio commented 6 years ago

how does this affect @JelleAalbers's S1 width matching (in waveform simulator summary note)?

JelleAalbers commented 6 years ago

Thanks for the study @zhut19. I used the simulation code in https://github.com/JelleAalbers/xeshape for my matching, which is slightly different than fax (but much faster, so waveform template matching in several dimensions is more practical). I tried to make them as similar as possible, but I guess some residual difference remains, causing this discrepancy. When I have time I'll try to compare the two in more detail and hopefully reconcile them. For now of course it's agreement between fax and data that counts, so this adjustment is very welcome.