MadGraph reweight technique

Asked by Yue Xu on 2019-10-04

Hello,

I generated the events with an external model which contains a heavy higgs ( hh ), and the process is:
  p p > z hh, z > l- l+, hh > j j j j

There are two parameters we can change. They are fw and fww that influence the coupling between heavy higgs and vector boson and the decay width of heavy higgs.
I would like to use MadGraph to reweight the generated events with a different setting of fw and fww.

For example, I generated events with fw=600 and fw=3600 (case A). Then, I want to reweight these events with fw=800 and fww=4800 (case B). The jet pt distribution of case A is similar with the jet pt distribution of case B. The cross-section of case B is larger than the cross-section of case A in theory.

I reweight the events by command:
        ./bin/madevent rewight run_01 -f

The setting of reweight_card.dat is ( the decay width of heavy higgs is always set to auto):
       launch --rwgt_name=fw800fww4800
      set DIM6COEFF 3 8.0000e+02
      set DIM6COEFF 4 4.8000e+03
      set DECAY 254 auto

Then, I got the following output information:
           INFO: central : 7.741888-112e-05
           INFO: fw800fww4800 : 7.0747137441e-05

It seems the reweighted cross-section isn't correct and not corresponding with the MG5 cross-section.

I also tried with different settings of fw and fww. I found that the reweighted cross-section is corresponding with the MG5 cross-section sometimes, but sometimes isn't.

How do these happen? How can I get a correct reweighted cross-section?

Thanks.

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Yue Xu
Solved:
2019-11-11
Last query:
2019-11-11
Last reply:
2019-10-30

This question was reopened

Hi,

If the width is dfferent in both cases then you can have large difference in the M_JJJJ distributions and therefore huge bias in the re-weighting method.

Another stuff that you might need to check is to de-activate the re-weighting by polarization.
But note that re-weighting is not expected to always work so like this I do not see any obvious problem here.

Cheers,

Olivier
> On 4 Oct 2019, at 14:22, Yue Xu <email address hidden> wrote:
>
> New question #684695 on MadGraph5_aMC@NLO:
> https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fanswers.launchpad.net%2Fmg5amcnlo%2F%2Bquestion%2F684695&amp;data=02%7C01%7C%7Cbd8aa81279a0478a233f08d748c5923a%7C7ab090d4fa2e4ecfbc7c4127b4d582ec%7C0%7C0%7C637057886171359966&amp;sdata=ICDv%2FDywrVUKlTIr0XnCwSH8eTc%2FQByM12d0606opOQ%3D&amp;reserved=0
>
> Hello,
>
> I generated the events with an external model which contains a heavy higgs ( hh ), and the process is:
> p p > z hh, z > l- l+, hh > j j j j
>
> There are two parameters we can change. They are fw and fww that influence the coupling between heavy higgs and vector boson and the decay width of heavy higgs.
> I would like to use MadGraph to reweight the generated events with a different setting of fw and fww.
>
> For example, I generated events with fw=600 and fw=3600 (case A). Then, I want to reweight these events with fw=800 and fww=4800 (case B). The jet pt distribution of case A is similar with the jet pt distribution of case B. The cross-section of case B is larger than the cross-section of case A in theory.
>
> I reweight the events by command:
> ./bin/madevent rewight run_01 -f
>
> The setting of reweight_card.dat is ( the decay width of heavy higgs is always set to auto):
> launch --rwgt_name=fw800fww4800
> set DIM6COEFF 3 8.0000e+02
> set DIM6COEFF 4 4.8000e+03
> set DECAY 254 auto
>
> Then, I got the following output information:
> INFO: central : 7.741888-112e-05
> INFO: fw800fww4800 : 7.0747137441e-05
>
> It seems the reweighted cross-section isn't corrected and not corresponding with the MG5 cross-section.
>
> I also tried with different settings of fw and fww. I found that the reweighted cross-section is corresponding with the MG5 cross-section sometimes, but sometimes isn't.
>
> How do these happen? How can I get a corrected reweighted cross-section?
>
> Thanks.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Yue Xu (yuexu) said : #2

Hi Olivier,

What else will influence the re-weighting method except the decay width ?

 If two signal points have a similar decay width and the pt distribution of vector boson is much different, will the re-weighting work ?

Cheers,
Yue

Full discussion is given in this paper: arxiv:1607.00763<http://arXiv.org/abs/arXiv:1607.00763>

But in short if
1) you have infinite statistic
2) your rewingting hyppothesis has a phase-space fully include in the original phase-space
(be carefull here with helicity by helicity re-weighting))
3) that the re-weighitng does not affect the leading color information of the events

Then you are sure to have the correct result.
Obviously the point 1. can never be achieved and this means that if your shape are too different (in ANY possible variable) then you can face either bias and/or huge theoretical error.

Cheers,

Olivier

On 8 Oct 2019, at 18:07, Yue Xu <<email address hidden><mailto:<email address hidden>>> wrote:

Question #684695 on MadGraph5_aMC@NLO changed:
https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fanswers.launchpad.net%2Fmg5amcnlo%2F%2Bquestion%2F684695&amp;data=02%7C01%7C%7C6042c1e6c19f4212b41408d74c09b110%7C7ab090d4fa2e4ecfbc7c4127b4d582ec%7C0%7C0%7C637061476819161778&amp;sdata=AOCpR%2BjTnNyfFlFdvoxCPGlfUZohgcfwY9WqM8QrsN8%3D&amp;reserved=0

   Status: Answered => Open

Yue Xu is still having a problem:
Hi Olivier,

What else will influence the re-weighting method except the decay width
?

If two signal points have a similar decay width and the pt distribution
of vector boson is much different, will the re-weighting work ?

Cheers,
Yue

--
You received this question notification because you are an answer
contact for MadGraph5_aMC@NLO.

Yue Xu (yuexu) said : #4

Hi,

I generated events with fw=0, fww=9000 and mass of heavy higgs is 900GeV. The decay width of heavy higgs is 32.2 GeV.
Then, I tried to reweight these events with fw=0 and fww=9000, keeping the parameter as same as before.

Here is the output result:

INFO: fw0fww9000 : 0.000171404534606 pb
INFO: central : 9.9899110089e-05 pb

Why does reweighting not work, even with the same parameters ?

Sounds weird, you can have small differences when you have negative weights (related to enhanced statistical error and how estimator are used) but in this case the difference sounds too big to be the case.

Can you send me the model? I will need it to investigate.

Cheers,

Olivier

Yue Xu (yuexu) said : #6

Hi,

Here is the link of model:
https://github.com/xuyue1231/MG5-heavy-Higgs/tree/master/SMwithHeavyScalarDim4Dim6_NoDecay_UFO
It contains both dim-4 and dim-6 operators of heavy higgs.

And the process p p > z hh, z > l- l+, hh > j j j j is here, including the setting of cards:
https://github.com/xuyue1231/MG5-heavy-Higgs/tree/master/rw_lljjjj

Cheers,
Yue

Hi,

I get the following:

INFO: Original cross-section: 9.509435e-05 +- 3.236154e-07 pb (cross-section from sum of weights: 9.5094353e-05)
INFO: Computed cross-section:
INFO: fw0fww9000 : 9.5094353e-05 +- 3.236154e-07 pb

But I see that you are using a very old version of the code. I would advise you to update since it looks like that you face a bug that was already solved.

Cheers,

Olivier

> On 29 Oct 2019, at 06:09, Yue Xu <email address hidden> wrote:
>
> Question #684695 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/684695
>
> Status: Answered => Open
>
> Yue Xu is still having a problem:
> Hi,
>
> Here is the link of model:
> https://github.com/xuyue1231/MG5-heavy-Higgs/tree/master/SMwithHeavyScalarDim4Dim6_NoDecay_UFO
> It contains both dim-4 and dim-6 operators of heavy higgs.
>
> And the process p p > z hh, z > l- l+, hh > j j j j is here, including the setting of cards:
> https://github.com/xuyue1231/MG5-heavy-Higgs/tree/master/rw_lljjjj
>
> Cheers,
> Yue
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Yue Xu (yuexu) said : #8

Thanks Olivier Mattelaer, that solved my question.

Yue Xu (yuexu) said : #9

Thanks a lot for the answer. When I use version 2.6.6, it works. And I still have a small question to ask you.

Could you tell me from what version this bug solved ?

Cheers,
Yue

I do not know that.

Cheers,

Olivier

Yue Xu (yuexu) said : #11

Hi Olivier,

Thanks a lot!

Cheers,
Yue