STAR HardProbes PWG

[Youqi Song <youqi.song AT yale.edu>]

Hi Nihar,

I assume these plots are approved now? I created a post just with the plots: https://drupal.star.bnl.gov/STAR/blog/youqi/Multidimensional-jet-substructure-measurement-unfolded-machine-learning-method-200-GeV-pp , so you could link them to the list of preliminary plots from HP PWG. Thanks!

Best,

Youqi

On Wed, Sep 28, 2022 at 12:54 PM Nihar Sahoo <nihar AT rcf.rhic.bnl.gov> wrote:

Hi Youqi,

Please find my replies inline.

On 2022-09-28 20:41, Youqi Song wrote:
> Hi Nihar,
>
> To implement -4% efficiency, we looped over each track in the
> embedding files and generated a random number from a uniform
> distribution from 0 to 1, and if the random number is greater than
> 0.96, we dropped that track when clustering jets.

Interesting, let's discuss this after HQ conference. For STAR
preliminary, it is Ok what you have.
Please remind us these topics in your next presentations.

>
> I lowered the y-axis limit and updated the figures on slides 19 and 20
> here:
> https://drupal.star.bnl.gov/STAR/system/files/prelim_request_092822.pdf

All your plots look great.

>
> They haven't announced the timetable yet, so we are not sure how long
> the talks need to be, but I can try to get a draft done by the end of
> the week.
Ok.


Cheers
Nihar

>
> Best,
> Youqi
>
> On Wed, Sep 28, 2022 at 2:11 AM Nihar Sahoo <nihar AT rcf.rhic.bnl.gov>
> wrote:
>
>> Hi Youqi,
>>
>> Please find my remaining question and comments inline.
>>
>> On 2022-09-26 21:52, Youqi Song wrote:
>>> Hi Nihar,
>>>
>>> Thanks for the comments. Here are the updated slides
>>>
>>
> https://drupal.star.bnl.gov/STAR/system/files/prelim_request_092622.pdf
>>>
>>> Yes, we can discuss the issue with smearing response after HQ.
>>>
>>> As for the tracking uncertainty, it might be difficult to try
>>> increasing it by 4%, since we are using official embedding and
>> can't
>>> add tracks back.
>>>
>> Not sure, if I understand your procedure correctly, can you inform
>> how
>> do you implement then -4% in embedding?
>>
>>> The ratio plot shown is MultiFold/RooUnfold, so I'm not sure what
>> else
>>> I can label the y-axis to be. Maybe it's more clear now that I
>>> combined the ratio plot with the main plot of fig 1 in slide 15.
>>
>> Ok, I understand now, I was thinking about the uncertainty band. But
>> it
>> is ok.and your plot looks nice now on slide#15.
>>
>> Can you please also lower y- axis for plots in slide#20 like in
>> slide15?
>>
>> Looking forward to seeing your HQ presentation draft.
>>
>> Cheers
>> Nihar
>>
>>>
>>> Best,
>>> Youqi
>>>
>>> On Mon, Sep 26, 2022 at 8:05 AM Nihar Sahoo
>> <nihar AT rcf.rhic.bnl.gov>
>>> wrote:
>>>
>>>> Hello Youqi,
>>>>
>>>> Please find my reply and queries inline.
>>>>
>>>> On 2022-09-25 23:02, Youqi Song wrote:
>>>>> Hi Nihar,
>>>>>
>>>>> Thanks for the feedback. I updated the slides here:
>>>>>
>>>>
>>>
>>
> https://drupal.star.bnl.gov/STAR/system/files/prelim_request_092522.pdf
>>>>>
>>>>> Response to unnumbered comments:
>>>>>
>>>>> Yes, I have three MCs, Herwig, Pythia8 and Pythia6, but what I
>> did
>>>>> earlier was just to use Pythia6 for the *misses*, I still did
>>>>> unfolding with each of the different MC prior shapes. (Now the
>>>> plots I
>>>>> updated on Friday have both the misses and the unfolded parts
>> with
>>>>> different priors). I plotted the fully corrected distributions
>>>> with
>>>>> the different priors together with other sources of sys
>>>> uncertainty in
>>>>> backup slide 1, and I just added a backup slide 2 with just the
>>>>> distributions due to prior shape variation.
>>>>
>>>> Ok, that's fine.
>>>>>
>>>>> Ok, I've added a slide after slide 15 to show the error
>> breakdown.
>>>>
>>>> I like this comparison plot that is in Slide#16. It looks like at
>>>> this
>>>> stage both Multifold and RooUnfold give roughly the same
>>>> uncertainties.
>>>> But we need to go through some other systematic uncertainties
>> (like
>>>> you
>>>> mentioned below) in future after HQ conf.
>>>>>
>>>>> For detector and generator level pT shape smearing, I think the
>>>> idea
>>>>> is to add some smearing to the response matrix itself. Maybe
>>>> others
>>>>> can correct me if I'm wrong.
>>>>>
>>>> Not sure if I understand this correctly, this needs a discussion.
>>>> Could
>>>> you please bring this topic up in your next presentation?
>>>> But for HQ, I think it is Ok, the systematic uncertainties you
>> have
>>>> for
>>>> STAR preliminary.
>>>>
>>>>> Response to numbered comments:
>>>>>
>>>>> 1. For the detector response systematics, there are hadronic
>>>>> correction, tower scale and tracking efficiency variation.
>>>>> For hadronic correction, I see how it makes sense to vary in
>> data
>>>> as
>>>>> well, but I learned that Raghav and Isaac didn't vary data for
>>>> their
>>>>> analyses, so I decided to go with how they did it. Also, I tried
>>>>> varying data as well, but it didn't have a large effect. See the
>>>>> comparison plot I added in slide 12.
>>>>
>>>> Good that you checked this for Hadronic correction.
>>>>
>>>>> For tower scale and tracking efficiency, I am not sure how we
>>>> would
>>>>> vary data. Right now I decreased the tracking efficiency in
>> Geant
>>>> by
>>>>> 4%, and kept data the same. Do you mean that we should drop 4%
>>>> tracks
>>>>> in data for this variation as well?
>>>>
>>>> No, we don't change anything in the data. We can only change in
>> the
>>>> tracking efficiency from embedding.
>>>> You mentioned that you only decreased 4%. Have you considered
>> also
>>>> increasing 4% in your systematic uncertainties ? Just to confirm.
>>>>
>>>>>
>>>>> 2. Great, I updated the figure in the slide.
>>>>
>>>> Good, Can you please lower your y-axis scale (say -0.01) so that
>> all
>>>> the
>>>> markers can be seen properly, particularly at large M region?
>>>> (in a plot, all data points should be presented/shown clearly)
>>>>
>>>>>
>>>>> 3. The star markers are MultiFold mass distributions divided by
>>>>> RooUnfold mass, so I'm not sure why we want a different style
>>>> marker?
>>>>
>>>> Suggesting because ratio star marker no need to be the same with
>>>> Multifold marker style. just for cosmetic.
>>>>
>>>>> The error bands are my unfolding systematics and RooUnfold
>>>> unfolding
>>>>> systematics added in quadrature, and then divided by RooUnfold
>>>> mass
>>>>> distribution.
>>>>
>>>> Can you please label your Y-axis accordingly without labeling as
>>>> "Ratio
>>>> with RooUnfold"?
>>>> It is not clear what you have written above. Is not it?
>>>> Please put pT range, "STAR Preliminary", and other info like your
>>>> slide#19.
>>>>
>>>>>
>>>>> 4. Great, agreed :)
>>>>>
>>>>> 5. I updated the legend to include "Detroit tune". I'm not sure
>> we
>>>>> want to push for the q vs g separation physics message. Maybe
>>>> Helen
>>>>> and Raghav can say more about this, but my understanding is that
>>>> the
>>>>> quark and gluon jets can have similar fragmentation patterns at
>>>> 200
>>>>> GeV and there can be ambiguity regarding e.g. if you want to
>> call
>>>> a
>>>>> jet initiated from g->qqbar splitting a quark jet or a gluon
>> jet,
>>>> so
>>>>> instead it might be more interesting to treat the jets with
>>>> different
>>>>> M and Q as having different fragmentation patterns.
>>>>>
>>>> Ok, that makes sense to me.
>>>>
>>>> Cheers
>>>> Nihar
>>>>
>>>>> Best,
>>>>> Youqi
>>>>>
>>>>> On Sun, Sep 25, 2022 at 8:23 AM Nihar Sahoo
>>>> <nihar AT rcf.rhic.bnl.gov>
>>>>> wrote:
>>>>>
>>>>>> Hello Youqi,
>>>>>>
>>>>>> Thank you for your update and this information.
>>>>>>
>>>>>> I have the following comment and questions on your new updated
>>>>>> results
>>>>>> (plots) and also your notes.
>>>>>>
>>>>>>> Regarding systematics,
>>>>>>> - You might have noticed that the systematic errors for
>>>> multifold
>>>>>> have
>>>>>>> gone up by a bit. This is because I realized that before, when
>>>>>> doing
>>>>>>> herwig and pythia8 shape variations, I was using pythia6 mass
>>>>>>> distribution for the misses. Now I changed it so that the
>> misses
>>>>>>> contribution
>>>>>>> for prior shape variation is also weighted by the mass ratio
>> of
>>>>>> herwig
>>>>>>> (pythia8) over pythia6.
>>>>>>
>>>>>> Good that you found out this.
>>>>>> If I understand correctly, you have three MCs: Herwig, Pythia8,
>>>> and
>>>>>> Pythia6. You used each of these MC mass distributions for the
>>>> mass
>>>>>> shape
>>>>>> variation for sys uncertainties. Earlier, you used Pythia6 for
>>>> all
>>>>>> these
>>>>>> cases, but now you corrected it with their respective mass
>>>> shapes.
>>>>>> Is
>>>>>> that correct?
>>>>>> Could you show (include in your backup) us those mass
>>>> distribution
>>>>>> variations using these three different MC priors? Curious to
>> see
>>>> the
>>>>>>
>>>>>> difference due to different fragmentations in the mass
>>>> distribution.
>>>>>>
>>>>>>
>>>>>>> - After talking with Isaac, I learned that unfolding
>> systematics
>>>>>> due to
>>>>>>> variation of iteration number and variation of prior shape
>> were
>>>>>> treated
>>>>>>> as correlated in previous analyses, while I originally had
>> these
>>>>>> added
>>>>>>> in quadrature. Now I also treat them as correlated by just
>>>> taking
>>>>>> the
>>>>>>> largest contribution to the unfolding systematic as the
>> overall
>>>>>>> unfolding systematic, and add it in quadrature with detector
>>>>>>> systematics.
>>>>>>
>>>>>> Could you please separate out I) statistical uncertainty, II)
>>>>>> Correlated
>>>>>> Sys, and III) Uncorrelated Sys in your plot ( using Style2 in
>>>>>> slide#16)
>>>>>> without adding in quadrature? Please use a smaller marker size
>> to
>>>>>> see
>>>>>> the stat. Error bar (even if it is small).
>>>>>>
>>>>>>> - Regarding the question whether there's anything still
>> missing
>>>>>> that's
>>>>>>> raised during the meeting yesterday, I forgot to mention that
>> we
>>>>>>> haven't included detector and generator level pT shape
>> smearing.
>>>>>> This
>>>>>>> is included in Raghav's and Isaac's analyses because they did
>> 1D
>>>>>>> reweighting, but we are not sure if it should be included for
>>>>>>> multidimensional unfolding.
>>>>>>
>>>>>> Hmm, Is not it that the total detector effects (tracking
>>>> efficiency
>>>>>> and
>>>>>> pT smearing) are considered using this multifold while you
>> create
>>>>>> something like the response matrix? What am I missing here?
>>>>>>
>>>>>> Comments on your slide#12 (from this email preliminary templet
>>>>>> slides)
>>>>>>
>>>>>> 1. Slide#12: As you have mentioned, you have varied only in the
>>>>>> embedding, not in the data. General practice is to vary in the
>>>> data.
>>>>>> But
>>>>>> not sure if that will be accurately translated to the variation
>>>> in
>>>>>> the
>>>>>> data. It is interesting. i) Then how do you use tracking
>>>> efficiency
>>>>>> (+-4%) variation in embedding to get systematic uncertainties?
>>>> ii)
>>>>>> Can
>>>>>> you please do a test where you apply the same variation in the
>>>> data
>>>>>> and
>>>>>> use the respective variation in the embedding and then check if
>>>> you
>>>>>> get
>>>>>> approximately the same sys. variation between the two cases
>>>> ?(just
>>>>>> for
>>>>>> one case which one has a bigger effect).
>>>>>>
>>>>>> 2. I like Style-4 (slide18) of figure-1 if you don't want to
>>>> shift
>>>>>> the
>>>>>> published results. In this case, I would suggest using an open
>>>> black
>>>>>>
>>>>>> star for published results and a filled red marker for your new
>>>>>> Multifold results. Then you plot your red filled star top on
>> the
>>>>>> published results (open back mark). In this case, you don't
>> need
>>>> to
>>>>>> shift the published results, and no confusion, and will look
>>>> good.
>>>>>>
>>>>>> 3. SLide19, Fig-1 ratio plot: In this plot, assuming
>> statistical
>>>>>> uncertainties cancel out. Are these bands only the ratio of
>>>>>> systematic
>>>>>> uncertainties? Can you use a separate marker style here in
>> order
>>>> to
>>>>>> avoid your mass distribution red star style?
>>>>>>
>>>>>> 4. SLide#21: Thank you for including raw distribution here. It
>>>> looks
>>>>>>
>>>>>> like there is a noticeable difference between raw and Multifold
>>>>>> levels.
>>>>>>
>>>>>> 5. Slide#23: Thank you for including PYTHIA8 curves. I like
>> this
>>>>>> plot
>>>>>> and the comparison. Pythia8 and Data are consistent. You could
>>>> point
>>>>>> out
>>>>>> that further study is ongoing to explore mass distribution
>>>> between q
>>>>>> vs.
>>>>>> g using this |Q| cut. This is the main physics of this |Q| cut
>>>> here.
>>>>>>
>>>>>> Right? (Can you label "PYTHiA8" as "PYTHIA8 Detroit tune" or
>>>> …STAR
>>>>>>
>>>>>> tune?)
>>>>>>
>>>>>> Cheers
>>>>>> Nihar
>>>>>>
>>>>>> On 2022-09-23 20:29, Youqi Song via Star-hp-l wrote:
>>>>>>> Hi all,
>>>>>>>
>>>>>>> I have updated my slides for the preliminary request at:
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
> https://drupal.star.bnl.gov/STAR/system/files/prelim_request_092222_0.pdf.
>>>>>>>
>>>>>>> A few things that I would like to point out:
>>>>>>>
>>>>>>> Regarding systematics,
>>>>>>> - You might have noticed that the systematic errors for
>>>> multifold
>>>>>> have
>>>>>>> gone up by a bit. This is because I realized that before, when
>>>>>> doing
>>>>>>> herwig and pythia8 shape variations, I was using pythia6 mass
>>>>>>> distribution for the misses. Now I changed it so that the
>> misses
>>>>>>> contribution for prior shape variation is also weighted by the
>>>>>> mass
>>>>>>> ratio of herwig (pythia8) over pythia6.
>>>>>>> - After talking with Isaac, I learned that unfolding
>> systematics
>>>>>> due
>>>>>>> to variation of iteration number and variation of prior shape
>>>> were
>>>>>>> treated as correlated in previous analyses, while I originally
>>>> had
>>>>>>> these added in quadrature. Now I also treat them as correlated
>>>> by
>>>>>> just
>>>>>>> taking the largest contribution to the unfolding systematic as
>>>> the
>>>>>>> overall unfolding systematic, and add it in quadrature with
>>>>>> detector
>>>>>>> systematics.
>>>>>>> - Regarding the question whether there's anything still
>> missing
>>>>>> that's
>>>>>>> raised during the meeting yesterday, I forgot to mention that
>> we
>>>>>>> haven't included detector and generator level pT shape
>> smearing.
>>>>>> This
>>>>>>> is included in Raghav's and Isaac's analyses because they did
>> 1D
>>>>>>> reweighting, but we are not sure if it should be included for
>>>>>>> multidimensional unfolding.
>>>>>>>
>>>>>>> Regarding plotting,
>>>>>>> - For fig. 1, I made the mass distribution plot with the same
>>>> data
>>>>>>> points in 4 different styles.- For fig. 1 ratio plot, the
>> error
>>>>>> band
>>>>>>> is now centered at 1 and is the quadrature of RooUnfold's and
>>>>>>> MultiFold's unfolding systematics, divided by the mean values
>>>> from
>>>>>>> RooUnfold.
>>>>>>> - For fig. 2, I also included a plot of the M vs Q correlation
>>>>>> with
>>>>>>> raw data before unfolding.
>>>>>>> - For fig. 3, I have pythia8 curves plotted together with my
>>>>>> unfolded
>>>>>>> mass distributions.
>>>>>>>
>>>>>>> Please let me know if you have any comments/suggestions.
>>>>>>>
>>>>>>> Best,
>>>>>>> Youqi
>>>>>>>
>>>>>>> On Wed, Sep 21, 2022 at 5:47 PM Youqi Song
>> <youqi.song AT yale.edu>
>>>>>>> wrote:
>>>>>>>
>>>>>>>> Hi all,
>>>>>>>>
>>>>>>>> I have uploaded my slides here:
>>>>>>>> https://drupal.star.bnl.gov/STAR/blog/youqi/multifold092222
>>>>>>>>
>>>>>>>> Best,
>>>>>>>> Youqi
>>>>>>>>
>>>>>>>> On Wed, Sep 21, 2022 at 5:04 PM Tong Liu via Star-hp-l
>>>>>>>> <star-hp-l AT lists.bnl.gov> wrote:
>>>>>>>>
>>>>>>>> Hi all,
>>>>>>>>
>>>>>>>> Please find my pdf in this post:
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
> https://drupal.star.bnl.gov/STAR/blog/tongliu/Tong-Lius-HP-PWG-updates
>>>>>>>>
>>>>>>>> Tong Liu
>>>>>>>> Ph.D. Student '2023
>>>>>>>> Physics Dept., Yale University
>>>>>>>>
>>>>>>>> Tel: 203-435-2130
>>>>>>>>
>>>>>>>> On Wed, Sep 21, 2022 at 3:29 AM Nihar Sahoo via Star-hp-l
>>>>>>>> <star-hp-l AT lists.bnl.gov> wrote:
>>>>>>>> Hello All,
>>>>>>>>
>>>>>>>> We got requests from Youqi and Tong to present their results
>>>> for
>>>>>> the
>>>>>>>> Hot
>>>>>>>> Quark conference.
>>>>>>>> And during Youqi's talk at the collaboration meeting, we did
>>>> not
>>>>>> get
>>>>>>>>
>>>>>>>> time to have Q&A.
>>>>>>>> So let's meet this week to discuss their updates.
>>>>>>>>
>>>>>>>> Youqi and Tong, can you please send link of your slides in
>> this
>>>>>>>> thread
>>>>>>>> again?
>>>>>>>> (I didn't get your previous emails, the reason I do not know)
>>>>>>>>
>>>>>>>> If anybody wants to discuss their results, please let us
>> know.
>>>>>>>>
>>>>>>>> HP-pwg weekly meeting Drupal page:
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
> https://drupal.star.bnl.gov/STAR/pwg/Hard-Probes/Weekly-HP-PWG-meeting
>>>>>>>>
>>>>>>>> Zoom Meeting link:
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
> https://bnl.zoomgov.com/j/1611419615?pwd=VW1hNm43ZDd5d2EvK2R4aEJsQ2ZNZz09
>> [1]
>>>> [1]
>>>>>> [1]
>>>>>>>> [1]
>>>>>>>>
>>>>>>>> Meeting ID: 161 141 9615
>>>>>>>> Passcode: 744968
>>>>>>>>
>>>>>>>> Thank you
>>>>>>>> Barbara, Yi, Sooraj, Nihar
>>>>>>>>
>>>>>>>> On 2022-09-19 22:13, Nihar Sahoo via Star-hp-l wrote:
>>>>>>>>> Hello All,
>>>>>>>>>
>>>>>>>>> Last week we discussed a lot at the collaboration meeting,
>> If
>>>>>>>> there is
>>>>>>>>> no urgent matter to discuss,  let's cancel this week's HP
>> pwg
>>>>>>>> meeting.
>>>>>>>>>
>>>>>>>>> Have a great week.
>>>>>>>>>
>>>>>>>>> Thank you
>>>>>>>>> Barbara, Yi, Sooraj, and Nihar
>>>>>>>>> _______________________________________________
>>>>>>>>> Star-hp-l mailing list
>>>>>>>>> Star-hp-l AT lists.bnl.gov
>>>>>>>>> https://lists.bnl.gov/mailman/listinfo/star-hp-l
>>>>>>>> _______________________________________________
>>>>>>>> Star-hp-l mailing list
>>>>>>>> Star-hp-l AT lists.bnl.gov
>>>>>>>> https://lists.bnl.gov/mailman/listinfo/star-hp-l
>>>>>>>> _______________________________________________
>>>>>>>> Star-hp-l mailing list
>>>>>>>> Star-hp-l AT lists.bnl.gov
>>>>>>>> https://lists.bnl.gov/mailman/listinfo/star-hp-l
>>>>>>>
>>>>>>>
>>>>>>> Links:
>>>>>>> ------
>>>>>>> [1]
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
> https://urldefense.com/v3/__https://bnl.zoomgov.com/j/1611419615?pwd=VW1hNm43ZDd5d2EvK2R4aEJsQ2ZNZz09__;!!P4SdNyxKAPE!HXAxDjpXwllezGXUg12_CP_CyiB1LboH0iAUfbzsPqj3eNb5aUboVO9QQZ6XzS2n6-OJ3TZo5Nq7ZITGfqCo0e94OA$
>>>>>>> _______________________________________________
>>>>>>> Star-hp-l mailing list
>>>>>>> Star-hp-l AT lists.bnl.gov
>>>>>>> https://lists.bnl.gov/mailman/listinfo/star-hp-l
>>>>>
>>>>>
>>>>> Links:
>>>>> ------
>>>>> [1]
>>>>>
>>>>
>>>
>>
> https://urldefense.com/v3/__https://bnl.zoomgov.com/j/1611419615?pwd=VW1hNm43ZDd5d2EvK2R4aEJsQ2ZNZz09__;!!P4SdNyxKAPE!BoY0IMYD87YlyC0E6bH2VCPQIHcRLzJj7X0OxGoAGqDB7vvV4xV5kA52ml71HdxJtHwEu26395xLJGHGX59wTJP5rw$
>>>
>>>
>>> Links:
>>> ------
>>> [1]
>>>
>>
> https://urldefense.com/v3/__https://bnl.zoomgov.com/j/1611419615?pwd=VW1hNm43ZDd5d2EvK2R4aEJsQ2ZNZz09__;!!P4SdNyxKAPE!AY99onlbPZAERQ8JTv73g9j81ASM172efJxmpy1l96C33ydc1qsfPOfY5pGMC1asn5Pe-WImnkYUnzek8vvo1KzX1A$
>
>
> Links:
> ------
> [1]
> https://urldefense.com/v3/__https://bnl.zoomgov.com/j/1611419615?pwd=VW1hNm43ZDd5d2EvK2R4aEJsQ2ZNZz09__;!!P4SdNyxKAPE!Cq5FUOmAj3PN2Z2A5doGpW0_GsQ-JMSRIqVT7tKFziK5UUn-rmtfReO_1IZvOXbuQNvpyjzPlnmfM4gk5dYkJ3QYFA$