Reweighting on quadratic terms of SMEFT gives wrong cross sections

Asked by xiao jie on 2020-07-30

Dear authors,

I use MG2.6.5 with SMEFT UFO model.

I want to reweight the quadratic term of same-sign WW VBS:
generate p p > l+ l+ vl vl j j QCD=0 SMHLOOP=0 NP=1 NP^2==1 @1
add process p p > l- l- vl~ vl~ j j QCD=0 SMHLOOP=0 NP=1 NP^2==1 @2

1) I active two operators Oqq3, Oqq31, their wilson coefficients: Cqq3=1, Cqq31=1 as starting point. Others are 0.

The reweight_card.dat is as following:
change helicity False
change rwgt_dir rwgt
launch --rwgt_name=all
set frblock 57 1
set frblock 58 1
launch --rwgt_name=cqq31
set frblock 57 1
set frblock 58 0
launch --rwgt_name=cqq31_1
set frblock 57 0
set frblock 58 1

I got cross sections for them:
INFO: Computed cross-section:
INFO: all : 0.621743879191 +- 0.00045123822203 pb
INFO: cqq31_1 : 6.95108318649 +- 1.65283393872 pb
INFO: cqq3_1 : 6.79616051861 +- 1.63204581137 pb

2) I also simulated quadratic terms independently for cqq3 and cqq31:
the cross sections are:
cqq31=1 : 0.1828pb
cqq3=1 : 0.1798pb

If you compare cross sections between 1) and 2) for cqq31=1 and cqq3=1. They are very different.
Do you know what's the problem?

Thanks,
Jie

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Last query:
2020-07-30
Last reply:
2020-07-30

Hi,

When you see this:

> INFO: all : 0.621743879191 +- 0.00045123822203 pb

> INFO: cqq31_1 : 6.95108318649 +- 1.65283393872 pb
> INFO: cqq3_1 : 6.79616051861 +- 1.63204581137 pb

You can see that the relative eror increased from
0.0007 to 0.24
If you add the fact that such error propagation is based on a gaussian model and that such model might not be valid in your case this leads to the conclusion that indeed you have an issue in your re-weighting.

Let's be clear issue does not mean bugs. The re-weighting procedure is not bullet proof method.
If you have the same exact domain of integrattion, it should always work for infinite statistics but
obviously we might not have enough statistics to be in that domain --and even if we do the interest of the method will completly disapears in that regime.--

My understanding is that your study shows that you can not make such re-weighting for that model/benchmark. If you want to understand why you need to look at distributions and uunderstand which variables is causing the issue.

Cheers,

Olivier

> On 30 Jul 2020, at 15:30, xiao jie <email address hidden> wrote:
>
> New question #692145 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/692145
>
> Dear authors,
>
> I use MG2.6.5 with SMEFT UFO model.
>
> I want to reweight the quadratic term of same-sign WW VBS:
> generate p p > l+ l+ vl vl j j QCD=0 SMHLOOP=0 NP=1 NP^2==1 @1
> add process p p > l- l- vl~ vl~ j j QCD=0 SMHLOOP=0 NP=1 NP^2==1 @2
>
> 1) I active two operators Oqq3, Oqq31, their wilson coefficients: Cqq3=1, Cqq31=1 as starting point. Others are 0.
>
> The reweight_card.dat is as following:
> change helicity False
> change rwgt_dir rwgt
> launch --rwgt_name=all
> set frblock 57 1
> set frblock 58 1
> launch --rwgt_name=cqq31
> set frblock 57 1
> set frblock 58 0
> launch --rwgt_name=cqq31_1
> set frblock 57 0
> set frblock 58 1
>
> I got cross sections for them:
> INFO: Computed cross-section:
> INFO: all : 0.621743879191 +- 0.00045123822203 pb
> INFO: cqq31_1 : 6.95108318649 +- 1.65283393872 pb
> INFO: cqq3_1 : 6.79616051861 +- 1.63204581137 pb
>
> 2) I also simulated quadratic terms independently for cqq3 and cqq31:
> the cross sections are:
> cqq31=1 : 0.1828pb
> cqq3=1 : 0.1798pb
>
> If you compare cross sections between 1) and 2) for cqq31=1 and cqq3=1. They are very different.
> Do you know what's the problem?
>
> Thanks,
> Jie
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

I should add that for interference term those issue are likely much deeper since you can have negative interference and therefore part of the phase-space with close to zero value, which leads to a divergence in the re-weigthing procedure if you do not have the two theory cancelling at the same phase-space point. This is likely your issue here.

Cheers,

Olivier

> On 30 Jul 2020, at 21:35, Olivier Mattelaer <email address hidden> wrote:
>
> Question #692145 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/692145
>
> Status: Open => Answered
>
> Olivier Mattelaer proposed the following answer:
> Hi,
>
> When you see this:
>
>> INFO: all : 0.621743879191 +- 0.00045123822203 pb
>
>> INFO: cqq31_1 : 6.95108318649 +- 1.65283393872 pb
>> INFO: cqq3_1 : 6.79616051861 +- 1.63204581137 pb
>
> You can see that the relative eror increased from
> 0.0007 to 0.24
> If you add the fact that such error propagation is based on a gaussian model and that such model might not be valid in your case this leads to the conclusion that indeed you have an issue in your re-weighting.
>
> Let's be clear issue does not mean bugs. The re-weighting procedure is not bullet proof method.
> If you have the same exact domain of integrattion, it should always work for infinite statistics but
> obviously we might not have enough statistics to be in that domain --and even if we do the interest of the method will completly disapears in that regime.--
>
> My understanding is that your study shows that you can not make such re-
> weighting for that model/benchmark. If you want to understand why you
> need to look at distributions and uunderstand which variables is causing
> the issue.
>
> Cheers,
>
> Olivier
>
>
>> On 30 Jul 2020, at 15:30, xiao jie <email address hidden> wrote:
>>
>> New question #692145 on MadGraph5_aMC@NLO:
>> https://answers.launchpad.net/mg5amcnlo/+question/692145
>>
>> Dear authors,
>>
>> I use MG2.6.5 with SMEFT UFO model.
>>
>> I want to reweight the quadratic term of same-sign WW VBS:
>> generate p p > l+ l+ vl vl j j QCD=0 SMHLOOP=0 NP=1 NP^2==1 @1
>> add process p p > l- l- vl~ vl~ j j QCD=0 SMHLOOP=0 NP=1 NP^2==1 @2
>>
>> 1) I active two operators Oqq3, Oqq31, their wilson coefficients: Cqq3=1, Cqq31=1 as starting point. Others are 0.
>>
>> The reweight_card.dat is as following:
>> change helicity False
>> change rwgt_dir rwgt
>> launch --rwgt_name=all
>> set frblock 57 1
>> set frblock 58 1
>> launch --rwgt_name=cqq31
>> set frblock 57 1
>> set frblock 58 0
>> launch --rwgt_name=cqq31_1
>> set frblock 57 0
>> set frblock 58 1
>>
>> I got cross sections for them:
>> INFO: Computed cross-section:
>> INFO: all : 0.621743879191 +- 0.00045123822203 pb
>> INFO: cqq31_1 : 6.95108318649 +- 1.65283393872 pb
>> INFO: cqq3_1 : 6.79616051861 +- 1.63204581137 pb
>>
>> 2) I also simulated quadratic terms independently for cqq3 and cqq31:
>> the cross sections are:
>> cqq31=1 : 0.1828pb
>> cqq3=1 : 0.1798pb
>>
>> If you compare cross sections between 1) and 2) for cqq31=1 and cqq3=1. They are very different.
>> Do you know what's the problem?
>>
>> Thanks,
>> Jie
>>
>> --
>> You received this question notification because you are an answer
>> contact for MadGraph5_aMC@NLO.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Can you help with this problem?

Provide an answer of your own, or ask xiao jie for more information if necessary.

To post a message you must log in.