Question about the cross-section uncertainty that madgraph generate

Asked by Olympia on 2018-04-04

Dear all,

I would like to ask you a question about the uncertainty that MadGraph gives as an output since it is not so clear to me.
In the step, that I am creating the GridPack, I got the following cross-section:

Cross-section : 0.0005401 +- 1.486e-06 pb

Is this uncertainty (+-1.486e-06) comming from the scale uncertainty of the following lines:

           'use_syst' : 'True',
            'sys_scalefact' : '1 0.5 2',
            'sys_pdf' : 'NNPDF23_lo_as_0130_qed'

and if not how exactly is computed?

Thank you in advance,
Olympia

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Last query:
2018-04-05
Last reply:
2018-04-10

Hi,

For gridpack generation, we do not provide theoretical uncertainty.
(this makes sense since for theoretical uncertainty we need to have lhe events which does not exists at
the time of the gridpack generation)

So this error is a pure statistical error related to the method of integration that we use.

Cheers,

Olivier

> On 4 Apr 2018, at 11:52, Olympia <email address hidden> wrote:
>
> New question #667547 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/667547
>
> Dear all,
>
> I would like to ask you a question about the uncertainty that MadGraph gives as an output since it is not so clear to me.
> In the step, that I am creating the GridPack, I got the following cross-section:
>
> Cross-section : 0.0005401 +- 1.486e-06 pb
>
> Is this uncertainty (+-1.486e-06) comming from the scale uncertainty of the following lines:
>
> 'use_syst' : 'True',
> 'sys_scalefact' : '1 0.5 2',
> 'sys_pdf' : 'NNPDF23_lo_as_0130_qed'
>
> and if not how exactly is computed?
>
> Thank you in advance,
> Olympia
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Olympia (olympiadartsi) said : #2

Hello Olivier,

Thank you for your answer. I was wondering if there is any way to achieve a
lower uncertainty in the production cross section. The reason we would need
that is that we want to compare the inclusive WZjj production after
subtracting the electroweak WZjj production and the QCD WZjj production
cross section to the cross section of the QCD-electroweak WZjj
interferences.

But the QCD WZjj, having almost two orders of magnitude higher cross
section that the interferences also carries an important absolute
statistical uncertainty, so we cannot conclude of the computed
interferences are constructive or destructive.

Thank you very much in advance for your help,

Olympia

*Olympia Dartsi *
phone @ LAPP: +33 4 50 09 1647, LAPP e-mail: <email address hidden>
phone @ CERN: +41 2 27 67 4250, CERN e-mail: <email address hidden>
skype: odartsi

On 4 April 2018 at 13:10, Olivier Mattelaer <
<email address hidden>> wrote:

> Your question #667547 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/667547
>
> Status: Open => Answered
>
> Olivier Mattelaer proposed the following answer:
> Hi,
>
> For gridpack generation, we do not provide theoretical uncertainty.
> (this makes sense since for theoretical uncertainty we need to have lhe
> events which does not exists at
> the time of the gridpack generation)
>
> So this error is a pure statistical error related to the method of
> integration that we use.
>
> Cheers,
>
> Olivier
>
> > On 4 Apr 2018, at 11:52, Olympia <email address hidden>
> wrote:
> >
> > New question #667547 on MadGraph5_aMC@NLO:
> > https://answers.launchpad.net/mg5amcnlo/+question/667547
> >
> > Dear all,
> >
> > I would like to ask you a question about the uncertainty that MadGraph
> gives as an output since it is not so clear to me.
> > In the step, that I am creating the GridPack, I got the following
> cross-section:
> >
> > Cross-section : 0.0005401 +- 1.486e-06 pb
> >
> > Is this uncertainty (+-1.486e-06) comming from the scale uncertainty of
> the following lines:
> >
> > 'use_syst' : 'True',
> > 'sys_scalefact' : '1 0.5 2',
> > 'sys_pdf' : 'NNPDF23_lo_as_0130_qed'
> >
> > and if not how exactly is computed?
> >
> > Thank you in advance,
> > Olympia
> >
> > --
> > You received this question notification because you are an answer
> > contact for MadGraph5_aMC@NLO.
>
> --
> If this answers your question, please go to the following page to let us
> know that it is solved:
> https://answers.launchpad.net/mg5amcnlo/+question/667547/+
> confirm?answer_id=0
>
> If you still need help, you can reply to this email or go to the
> following page to enter your feedback:
> https://answers.launchpad.net/mg5amcnlo/+question/667547
>
> You received this question notification because you asked the question.
>

Hi,

You can in principle reduce the statistical uncertainty as much as you want.
But since you are typically limited by scale uncertainty they are typically no point to want to reduce such error that much.

For the interference case, note that madgraph can compute (LO) interference term.

the syntax for that is the following
generate p p > W Z j j QED^2==4
(If I understand correctly the interference term that you are looking for)

Note that
1) The ^2 indicates that the restriction of the coupling should be handle after the squaring of the diagram.
2) It is advise to NOT use the default dynamical scale for interference computation (you should rather use HT/2 in such case)
3) You can not decay the W/Z within MG5_aMC (same in MadSpin)

Cheers,

Olivier

> On 5 Apr 2018, at 17:43, Olympia <email address hidden> wrote:
>
> Question #667547 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/667547
>
> Status: Answered => Open
>
> Olympia is still having a problem:
> Hello Olivier,
>
> Thank you for your answer. I was wondering if there is any way to achieve a
> lower uncertainty in the production cross section. The reason we would need
> that is that we want to compare the inclusive WZjj production after
> subtracting the electroweak WZjj production and the QCD WZjj production
> cross section to the cross section of the QCD-electroweak WZjj
> interferences.
>
> But the QCD WZjj, having almost two orders of magnitude higher cross
> section that the interferences also carries an important absolute
> statistical uncertainty, so we cannot conclude of the computed
> interferences are constructive or destructive.
>
> Thank you very much in advance for your help,
>
> Olympia
>
> *Olympia Dartsi *
> phone @ LAPP: +33 4 50 09 1647, LAPP e-mail: <email address hidden>
> phone @ CERN: +41 2 27 67 4250, CERN e-mail: <email address hidden>
> skype: odartsi
>
> On 4 April 2018 at 13:10, Olivier Mattelaer <
> <email address hidden>> wrote:
>
>> Your question #667547 on MadGraph5_aMC@NLO changed:
>> https://answers.launchpad.net/mg5amcnlo/+question/667547
>>
>> Status: Open => Answered
>>
>> Olivier Mattelaer proposed the following answer:
>> Hi,
>>
>> For gridpack generation, we do not provide theoretical uncertainty.
>> (this makes sense since for theoretical uncertainty we need to have lhe
>> events which does not exists at
>> the time of the gridpack generation)
>>
>> So this error is a pure statistical error related to the method of
>> integration that we use.
>>
>> Cheers,
>>
>> Olivier
>>
>>> On 4 Apr 2018, at 11:52, Olympia <email address hidden>
>> wrote:
>>>
>>> New question #667547 on MadGraph5_aMC@NLO:
>>> https://answers.launchpad.net/mg5amcnlo/+question/667547
>>>
>>> Dear all,
>>>
>>> I would like to ask you a question about the uncertainty that MadGraph
>> gives as an output since it is not so clear to me.
>>> In the step, that I am creating the GridPack, I got the following
>> cross-section:
>>>
>>> Cross-section : 0.0005401 +- 1.486e-06 pb
>>>
>>> Is this uncertainty (+-1.486e-06) comming from the scale uncertainty of
>> the following lines:
>>>
>>> 'use_syst' : 'True',
>>> 'sys_scalefact' : '1 0.5 2',
>>> 'sys_pdf' : 'NNPDF23_lo_as_0130_qed'
>>>
>>> and if not how exactly is computed?
>>>
>>> Thank you in advance,
>>> Olympia
>>>
>>> --
>>> You received this question notification because you are an answer
>>> contact for MadGraph5_aMC@NLO.
>>
>> --
>> If this answers your question, please go to the following page to let us
>> know that it is solved:
>> https://answers.launchpad.net/mg5amcnlo/+question/667547/+
>> confirm?answer_id=0
>>
>> If you still need help, you can reply to this email or go to the
>> following page to enter your feedback:
>> https://answers.launchpad.net/mg5amcnlo/+question/667547
>>
>> You received this question notification because you asked the question.
>>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Olympia (olympiadartsi) said : #4

Hi Olivier,

Thank you very much for your answer.
 Indeed I computed the interference term only and I'm getting a cross section (for the interference only) which seems reasonable to me, but how can I know if this interference is contructive or destructive?
 I do not have to check the sign of (Inclusive-QD-EW) for that?

Also: you say that I can reduce the uncertainty of the first step as much as I want. But how can I do that (as in the first step I don't define any number of events)?
Then I understand that in the second step, as I define use_syst to be true I will anyway have a higher uncertainty. But if I don't want to take into account the scale uncertainty I can also it out to false, right?

Thank you very much again!
Olympia

> Thank you very much for your answer.
> Indeed I computed the interference term only and I'm getting a cross section (for the interference only) which seems reasonable to me, but how can I know if this interference is contructive or destructive?
> I do not have to check the sign of (Inclusive-QD-EW) for that?

If you cross-section is positive, then your interference is (most of the time) constructive
if your cross-section is negative then your interference is (most of the time) negative.
You can look at the weight distributions of the events to see how many of them are positive/negative to see if you have a mix
of constructive/destructive interference.

> Also: you say that I can reduce the uncertainty of the first step as much as I want. But how can I do that (as in the first step I don't define any number of events)?

Gridpack generation is not the most convenient method to reduce that number actually. Because you can only target the precision of each channel of integration independently which will lead to huge waste of cpu time.

I would therefore recommend that you do not use gridpack generation but standard mode.
There you can define your number of events directly (and if this number starts to be too big, you can use the multi_run method, this method will automatically combine the various cross-section estimator in a single one)

> Then I understand that in the second step, as I define use_syst to be true I will anyway have a higher uncertainty. But if I don't want to take into account the scale uncertainty I can also it out to false, right?

You can always do whatever you want.The question is obviously how much can you trust your result.
And this really depends of the observable that you use.
I'm typically refusing paper that I review if they do not include such uncertainties.

Cheers,

Olivier

> On 10 Apr 2018, at 10:04, Olympia <email address hidden> wrote:
>
> Question #667547 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/667547
>
> Olympia posted a new comment:
> Hi Olivier,
>
> Thank you very much for your answer.
> Indeed I computed the interference term only and I'm getting a cross section (for the interference only) which seems reasonable to me, but how can I know if this interference is contructive or destructive?
> I do not have to check the sign of (Inclusive-QD-EW) for that?
>
> Also: you say that I can reduce the uncertainty of the first step as much as I want. But how can I do that (as in the first step I don't define any number of events)?
> Then I understand that in the second step, as I define use_syst to be true I will anyway have a higher uncertainty. But if I don't want to take into account the scale uncertainty I can also it out to false, right?
>
> Thank you very much again!
> Olympia
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Olympia (olympiadartsi) said : #6

Dear Olivier,

thanks a lot, that was very helpfull.

Cheers,
Olympia

Can you help with this problem?

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

To post a message you must log in.