STAR Correlations and Fluctuations PWG

[Hanna Zbroszczyk <hanna.zbroszczyk AT pw.edu.pl>]

Hi Sneha and All,

As mentioned before several times, if we want to make progress, we need much more discussion focused on the physics. I hope a paper is something you are considering as a goal for this project. Nu brought up several recurring issues, from understanding the concept of Levy walks to clarifying the physics behind this analysis, which remains unclear to me as well. The quality of the fit is undoubtedly something critical, to rely on for your findings.

We agreed that you would present preliminary results on Levy fits during WPCF, as this workshop provides an excellent opportunity to discuss the findings with experts. At the STAR Collaboration meeting two weeks ago, Mate presented on the Levy analysis, and I believe that after his talk it is clear we need further steps to move the analysis forward, especially if you are planning to submit a QM abstract. We need to ensure that STAR is prepared with a clear physics message by that time, and that it integrates with all previous efforts. 

I am looking forward to ongoing discussions,

Thanks,

Hanna

Prof. Hanna Zbroszczyk, PhD, DSc, Eng.

E-mail: hanna.zbroszczyk AT pw.edu.pl
Tel: +48 22 234 5851 (office)

Address:
Warsaw University of Technology
Faculty of Physics
Nuclear Physics Division
Koszykowa 75
Office: 117b (via 115)
00-662 Warsaw, Poland

Wiadomość napisana przez Nu Xu <nxu AT lbl.gov> w dniu 03.11.2024, o godz. 00:00:

Dear Dear  Sneha,

Let me make the last comment on this issue:
For p-value, when it is less than 0.05, the fit result is considered to be ok. The lower plot on slide 14 shows, however, p-values are larger than 0.05 for mT>=0.5 GeV from both alpha =2 and Levy fits. To me it is difficult to see the difference, especially if one makes a plot with Chi^2/def.

I sign off the talk for now. But one should really think how to make the discussion regarding the goodness of the fit. I am sure that we will again discuss about this when we write the paper.

Best regards,
Nu

On Nov 2, 2024, at 9:52 AM, Bhosale Sneha <bhosales AT staff.elte.hu> wrote:

Dear Nu,
Thank you for the email. I now made clear that this is in fact indeed on the slide and put titles with large font to the chi^2 difference and p-value plots, on slide 14.

The updated version is now available here for your review: (version 4)     https://drupal.star.bnl.gov/STAR/node/69661

Thank you.
Best regards,
SnehaFrom: Nu Xu <nxu AT lbl.gov>
Sent: Saturday, November 2, 2024 5:35 PM
To: Bhosale Sneha <bhosales AT staff.elte.hu>
Cc: Nu Xu <nxu AT lbl.gov>; STAR Correlations and Fluctuations PWG <star-cf-l AT lists.bnl.gov>
Subject: Re: [[Star-cf-l] ] WPCF 2024 talk
Dear  Sneha,
Is it possible to make it clear that Slide 14 is, infact, a displace of p-vale?
At the moment, as far as I can see, it is either (chia^2_apha =2 - chir^2) vs. mt or Count. Lev. vs.mt.?????!!!

Regards,
Nu

On Nov 2, 2024, at 7:06 AM, Bhosale Sneha <bhosales AT staff.elte.hu> wrote:

Dear Nu,
Thank you so much for the email. Here are my responses/explanation for you comments;
1) Plot The bottom right plot in slide 14 already shows the confidence levels (p-values) that you requested. This plot compares two cases:
Red triangles: fits with free α
Black triangles: fits with fixed α = 2.0 (Gaussian case)
2) The y-axis shows the confidence levels on a logarithmic scale from 10^-7 to 1, clearly demonstrating the several orders of magnitude improvement mentioned in the bullet points when comparing free α fits to Gaussian fits. This plot directly addresses the need for experimental evidence of the improved confidence levels rather than just stating the claim. Also, chi2/NDF difference plot is on top of the confidence level plot.
3)Slide 11: Yes, all the parameters are there in traditional gaussian fits
Thank you.
Best regards,
Sneha

From: Nu Xu <nxu AT lbl.gov>
Sent: Saturday, November 2, 2024 2:27 PM
To: Bhosale Sneha <bhosales AT staff.elte.hu>
Cc: Nu Xu <nxu AT lbl.gov>; STAR Correlations and Fluctuations PWG <star-cf-l AT lists.bnl.gov>
Subject: Re: [[Star-cf-l] ] WPCF 2024 talk
Dear Sneha,

Thank you for the updated version.
Here are my comment/suggestions on the v3:
(1) slide 14 - "Confidence levels (p-values) improve by 1-3 orders of magnitude with free 𝜶”: show a plot or remove the claim;
(2) "(ii): for all these fits, chi^2/NDF is approximately the same, around 1. It is however not a good number to estimate goodness-of-fit, as for large NDF (number of degrees of freedom) values very small changes in chi^2 can lead to large changes in probability. Thus, it is more important to show chi^2, but even more importantly, the probability (p-value or confidence level). “  So show a figure with p-values. Do NOT just claim, show the experimental figures!!!
(3) slide 11: are parameters already in the ’transitional fits with alpha=2”???

Regards,
Nu

On Nov 2, 2024, at 2:49 AM, Bhosale Sneha <bhosales AT staff.elte.hu> wrote:

Dear Nu,
Thank you so much for the email. Since my talk is on 5th of November 2024(Tuesday) I mostly removed the things you asked for and here are my responses/explanation for you comments;
1) A slide could be added to explain the physics meaning but given that now the talk is very close (Tuesday), I cannot do anything else but just remove these mentions. I do think the concept is important but it would probably need discussions within the PWG how the explanatory slides should look like, so for the sake of time, I removed these statements.
2) For the first point (i), "Lévy-scale R" is the size parameter of the Lévy distribution, meaning the size of the source. For the second point (ii): for all these fits, chi^2/NDF is approximately the same, around 1. It is however not a good number to estimate goodness-of-fit, as for large NDF (number of degrees of freedom) values very small changes in chi^2 can lead to large changes in probability. Thus, it is more important to show chi^2, but even more importantly, the probability (p-value or confidence level). Nevertheless, I now wrote on the slide that chi^2/NDF is approximately 1 for all fits.
Actually, the number of parameters differs just by one, alpha is the only additional parameters. More accurate description can be claimed because of the difference in the goodness-of-fit. But for the sake of reaching a good consensus, I removed this wording.
The updated version is now available here for your: (version 3)    https://drupal.star.bnl.gov/STAR/node/69661
Thank you again.
Best regards,
Sneha

From: Nu Xu <nxu AT lbl.gov>
Sent: Saturday, November 2, 2024 7:46 AM
To: Bhosale Sneha <bhosales AT staff.elte.hu>
Cc: Nu Xu <nxu AT lbl.gov>; STAR Correlations and Fluctuations PWG <star-cf-l AT lists.bnl.gov>
Subject: Re: [[Star-cf-l] ] WPCF 2024 talk
Dear Sneha,

Thanks for the updated presentation. I do have trouble with some claims shown in the present slides:

1) The term Levy Walk mentioned in slides 6, 13, 15, and 17.
We throw the term to audiences without explanations, it is very confusion. For this talk if the concept is needed, then provide the physics meaning. It it is not needed for this talk, take them all out of the presentation.

2) slide 14: (i) What does it mean by “Levy-scaled R”? (ii) right two plots: are these the chi^2/per degree of freedom? If it is, please note them so, for example, chi^2/pdf.
If I understand, there are many more parameters in the so-called Levy-fit, including epsilon_o, epsilon_s, epsilon_l?

A fit with many more parameters, in this case is the Levy-fit, can we really claim “… more accurate shape and size” shown on slide 5?

Best regards,
Nu

On Nov 1, 2024, at 3:24 AM, Bhosale Sneha <bhosales AT staff.elte.hu> wrote:

Dear Nu,

I wanted to follow up on my response to your helpful feedback on my presentation. I have incorporated the suggested changes and provided explanations where needed, including adjustments on slides 10, 11, 14, and 15, along with the reference addition. My responses to your suggestions can be found in below email.

The updated version is now available here for your review: https://drupal.star.bnl.gov/STAR/node/69661

Thank you once again for your valuable insights, and I look forward to any further suggestions you may have.

Best regards,
Sneha



Get Outlook for AndroidFrom: star-cf-l-request AT lists.bnl.gov <star-cf-l-request AT lists.bnl.gov> on behalf of Bhosale Sneha <bhosales AT staff.elte.hu>
Sent: Wednesday, October 30, 2024 4:58:25 PM
To: STAR Correlations and Fluctuations PWG <star-cf-l AT lists.bnl.gov>
Cc: Nu Xu <nxu AT lbl.gov>
Subject: Re: [[Star-cf-l] ] WPCF 2024 talk
Dear Nu,
Thank you so much for he feedback following is my response;
   • Slide 11:  I incorporated your suggestions, and have only one nozero digit in all of the uncertainties, and rounded the parameter values accordingly.
   • Slide 11: Thank you for pointing it out, it surely needs an explanation. I added the explanation of these parameters to slide 10 where they are discussed in last bullet point; let me also mention here that this essentially handles all residual, non-femtoscopic background, coming from energy and momentum conservation, resonance decays, bulk flow phenomena, minijets.
   • Slide 14: I agree. I now include the chi^2 difference and the p-value (confidence level) for both fits. The reference has also been added for published data.
   • Slide 15: Agree, it does need a bit explanation. Now I added 3 bullet points as an explanation.
I have incorporated your suggestion in version 2. Please have a look.
https://drupal.star.bnl.gov/STAR/node/69661
Thank you.
Best regards,
Sneha

From: star-cf-l-request AT lists.bnl.gov <star-cf-l-request AT lists.bnl.gov> on behalf of Nu Xu <nxu AT lbl.gov>
Sent: Wednesday, October 30, 2024 12:17 AM
To: STAR Correlations and Fluctuations PWG <star-cf-l AT lists.bnl.gov>
Cc: Nu Xu <nxu AT lbl.gov>
Subject: Re: [[Star-cf-l] ] WPCF 2024 talk
Hi Sneha,

Thank you for the draft for WPCF2025. I have the following suggestions for your consideration:
1) slide 11 - middle plot: those numbers listed there are too precise, perhaps meaningless. For example, lambda could be 0.99+-0.05, R_out could be 6.7+-0.2 fm and so on;
2) slide 11: what are those epsilons? Why do we need them? Any physics meanings of these parameters?

3)  slide 14 - last three bullets: instead of these claims, can we simply list the fitting results of chi^2/pdf from both alpha=2 and free alpha fits? For the alpha =2 data, a reference should be provided.

4) slide 15: what is "Levy walk”? One should explain it in the talk, not just point to another talk in my view;

That is all for now.
Best regards,

Nu

On Oct 29, 2024, at 5:26 AM, Bhosale Sneha <bhosales AT staff.elte.hu> wrote:

Dear All,
I would like to let you know that the WPCF conference agenda has been released, and my talk is scheduled for the 5th of November 2024, a week from now. I would greatly appreciate if you could review my contribution before the 4th of November. Your feedback and suggestions would be invaluable in improving the presentation.

https://drupal.star.bnl.gov/STAR/node/69661

Thank you in advance for your time and support.

Best regards,
SnehaFrom: star-cf-l-request AT lists.bnl.gov <star-cf-l-request AT lists.bnl.gov> on behalf of Bhosale Sneha <bhosales AT staff.elte.hu>
Sent: Wednesday, October 23, 2024 4:51 PM
To: STAR Correlations and Fluctuations PWG <star-cf-l AT lists.bnl.gov> <star-cf-l AT lists.bnl.gov>
Subject: [[Star-cf-l] ] WPCF 2024 talk
Dear All,
Please have a look at my contribution for WPCF2024 talk and let me know your views.

https://drupal.star.bnl.gov/STAR/node/69661

Thank you very much for considering my request.

Best regards,
Sneha