Remove S2 tail cut from calibration modes #88
Conversation
trailing whitespace
There was a problem hiding this comment.
This was done in #86 by @sreichard, but I requested to remove it from there to have a separate PR like this.
| @@ -52,6 +52,10 @@ class LowEnergyRn220(AllEnergy): | |||
|
|
|||
| def __init__(self): | |||
| AllEnergy.__init__(self) | |||
|
|
|||
| # S2Tails not used in calibration modes | |||
| self.lichen_list.pop(8) | |||
There was a problem hiding this comment.
Why not just delete the cut from the AllEnergy list above (like in #86)?
There was a problem hiding this comment.
We could. Technically it may also belong in "AllEnergy" since it should work the same at all energy scales.
There was a problem hiding this comment.
Ah I see sorry, missed that this was in LowEnergyRn220.
Maybe should move below:
self.lichen_list[4] = S2SingleScatterSimple()
in case future people pop more and affect the indexing of the cuts below.
|
Why do we need to remove this cut from the list? Can you remind me by adding a note. Same holds for pre-S2 junk cut. |
|
@skazama's note reported a loss of acceptance in calibration data, I guess because the cut was ultimately tuned for background data. But if I remember correctly, @coderdj @ramirezdiego, in e.g. SR0, you evaluated this for AmBe, so couldn't you also provide a different tuning for each source? |
|
|
|
@pdeperio the reason this was tuned first on AmBe in SR0 was because I hadn't realized yet we could use blinded background for it as long as we didn't include cs2 or make any controversial spectra. Once that method got established there was no reason to use AmBe anymore and it was dropped. I think it's not worth the time to tune this for each source. You can apply it as-is if you only want very clean events, but this should be done knowingly. |
|
Sorry, forgot to reply. Both cuts are very useful to validate the AC analysis, especially pre-S2 junk. I will update the test to see if we can have a reliable prediction without them. |
|
@feigaodm I don't understand. Why do you need to validate AC in calibration modes? This will get super tricky because of the correlation effect with tails of big S2s. With the tail cut in background mode lone S1/S2 should not come from an S2 tail but from something else. |
|
@coderdj We need AC model in Rn220 calibration data when Qing et.al performs the band fitting, hence AC model must be validated on calibration data as well. To do this, we need pre-S2 junk cut, otherwise there will be a lot of junk events which can not be modeled by AC analysis. This is especially true for events with s2 below the threshold of 200PE, junks from previouse event can bias the analysis a lot. |
|
I guess it's better to modify the cuts instead of removing them. @pdeperio The cuts has been considered in both prediction and validation stage so there is no mis-match due to those cuts. |
|
Yeah, I agree. I talked to @skazama and confirm that the acceptance loss is very bad for NG data only. For Rn220 data, I don't think the effect is very big, maybe 10-20% level. |
|
From these plots, AmBe and NG and Rn220 and BG, the acceptance loss for Rn220 is just 12% for this cut. For AmBe, it is 13% loss. |
We really need to do this