cross-sections of combined processes

Asked by Juhi Dutta on 2018-08-20

Dear experts,

I am comparing the cross-sections of two processes after generating them first as individual processes and then combining them in a single run.

For eg,

generate p p > t t~
output ttbar

generate p p > w+ w-
output ww

and

generate p p > t t~ @1
add process p p > w+ w- @2
output combined_proc

We compare the cross-sections of these processes for three different number of events, 10,1000 and 10000 keeping the same iseed value. However the cross-sections for ttbar, ww and combined_proc differ by upto 11 % for w+w- pair production whereas upto 0.5% for tt~ pair production in each case. I am unsure as to why this discrepancy occurs especially for ww where it is nearly 11 %.

I was wondering whether for the combined processes, could it be that the number of events for computing the cross-section of the combined process and not the individual process is 1000 events during survey whereas for the processes generated separately, survey uses 1000 events and hence resulting in better accuracy in the results?

Your advice in this regard will be very helpful to understand this issue.

Thanking You,

Best Regards,

Juhi

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Last query:
2018-08-23
Last reply:
2018-08-23

Hi,

> I was wondering whether for the combined processes, could it be that the number of events for computing the cross-section of the combined process and not the individual process is 1000 events during survey whereas for the processes generated separately, survey uses 1000 events and hence resulting in better accuracy in the results?

For the survey stage, the number of events is fixed per process but we only ask a 20% accuracy at that stage.
For the refine stage, the number of events depends on the total number of events that you request (in the run_card) and therefore the number of events required depends of the relative weight of the cross-section between the different process.

Cheers,

 Olivier

> On 20 Aug 2018, at 08:37, Juhi Dutta <email address hidden> wrote:
>
> New question #672506 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/672506
>
> Dear experts,
>
> I am comparing the cross-sections of two processes after generating them first as individual processes and then combining them in a single run.
>
> For eg,
>
> generate p p > t t~
> output ttbar
>
> generate p p > w+ w-
> output ww
>
> and
>
> generate p p > t t~ @1
> add process p p > w+ w- @2
> output combined_proc
>
> We compare the cross-sections of these processes for three different number of events, 10,1000 and 10000 keeping the same iseed value. However the cross-sections for ttbar, ww and combined_proc differ by upto 11 % for w+w- pair production whereas upto 0.5% for tt~ pair production in each case. I am unsure as to why this discrepancy occurs especially for ww where it is nearly 11 %.
>
> I was wondering whether for the combined processes, could it be that the number of events for computing the cross-section of the combined process and not the individual process is 1000 events during survey whereas for the processes generated separately, survey uses 1000 events and hence resulting in better accuracy in the results?
>
> Your advice in this regard will be very helpful to understand this issue.
>
> Thanking You,
>
> Best Regards,
>
> Juhi
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Juhi Dutta (dutta-juhi91) said : #2

Dear Sir,

Thank you for the prompt reply. Sorry but I still have a remaining query.
Could you please explain the difference in the survey and refine state
little more in detail and how the cross-section is finalized?

Best Regards,

Juhi

On Mon, Aug 20, 2018 at 1:32 PM Olivier Mattelaer <
<email address hidden>> wrote:

> Your question #672506 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/672506
>
> Status: Open => Answered
>
> Olivier Mattelaer proposed the following answer:
> Hi,
>
> > I was wondering whether for the combined processes, could it be that
> the number of events for computing the cross-section of the combined
> process and not the individual process is 1000 events during survey
> whereas for the processes generated separately, survey uses 1000 events
> and hence resulting in better accuracy in the results?
>
> For the survey stage, the number of events is fixed per process but we
> only ask a 20% accuracy at that stage.
> For the refine stage, the number of events depends on the total number of
> events that you request (in the run_card) and therefore the number of
> events required depends of the relative weight of the cross-section between
> the different process.
>
> Cheers,
>
> Olivier
>
> > On 20 Aug 2018, at 08:37, Juhi Dutta <
> <email address hidden>> wrote:
> >
> > New question #672506 on MadGraph5_aMC@NLO:
> > https://answers.launchpad.net/mg5amcnlo/+question/672506
> >
> > Dear experts,
> >
> > I am comparing the cross-sections of two processes after generating them
> first as individual processes and then combining them in a single run.
> >
> > For eg,
> >
> > generate p p > t t~
> > output ttbar
> >
> > generate p p > w+ w-
> > output ww
> >
> > and
> >
> > generate p p > t t~ @1
> > add process p p > w+ w- @2
> > output combined_proc
> >
> > We compare the cross-sections of these processes for three different
> number of events, 10,1000 and 10000 keeping the same iseed value. However
> the cross-sections for ttbar, ww and combined_proc differ by upto 11 % for
> w+w- pair production whereas upto 0.5% for tt~ pair production in each
> case. I am unsure as to why this discrepancy occurs especially for ww where
> it is nearly 11 %.
> >
> > I was wondering whether for the combined processes, could it be that the
> number of events for computing the cross-section of the combined process
> and not the individual process is 1000 events during survey whereas for the
> processes generated separately, survey uses 1000 events and hence resulting
> in better accuracy in the results?
> >
> > Your advice in this regard will be very helpful to understand this
> issue.
> >
> > Thanking You,
> >
> > Best Regards,
> >
> > Juhi
> >
> > --
> > 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/672506/+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/672506
>
> You received this question notification because you asked the question.
>

They are no difference in itself
They run the same code it just the stop condition which is different

For survey we stop at 20% accuracy
For the refine we stop when we have generated enough events if the survey generated enough then this stage is skipped

Cheers

Get Outlook for iOS<https://aka.ms/o0ukef>
________________________________
From: <email address hidden> <email address hidden> on behalf of Juhi Dutta <email address hidden>
Sent: Monday, August 20, 2018 10:23:05 AM
To: Olivier Mattelaer
Subject: Re: [Question #672506]: cross-sections of combined processes

Question #672506 on MadGraph5_aMC@NLO changed:
https://answers.launchpad.net/mg5amcnlo/+question/672506

    Status: Answered => Open

Juhi Dutta is still having a problem:
Dear Sir,

Thank you for the prompt reply. Sorry but I still have a remaining query.
Could you please explain the difference in the survey and refine state
little more in detail and how the cross-section is finalized?

Best Regards,

Juhi

On Mon, Aug 20, 2018 at 1:32 PM Olivier Mattelaer <
<email address hidden>> wrote:

> Your question #672506 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/672506
>
> Status: Open => Answered
>
> Olivier Mattelaer proposed the following answer:
> Hi,
>
> > I was wondering whether for the combined processes, could it be that
> the number of events for computing the cross-section of the combined
> process and not the individual process is 1000 events during survey
> whereas for the processes generated separately, survey uses 1000 events
> and hence resulting in better accuracy in the results?
>
> For the survey stage, the number of events is fixed per process but we
> only ask a 20% accuracy at that stage.
> For the refine stage, the number of events depends on the total number of
> events that you request (in the run_card) and therefore the number of
> events required depends of the relative weight of the cross-section between
> the different process.
>
> Cheers,
>
> Olivier
>
> > On 20 Aug 2018, at 08:37, Juhi Dutta <
> <email address hidden>> wrote:
> >
> > New question #672506 on MadGraph5_aMC@NLO:
> > https://answers.launchpad.net/mg5amcnlo/+question/672506
> >
> > Dear experts,
> >
> > I am comparing the cross-sections of two processes after generating them
> first as individual processes and then combining them in a single run.
> >
> > For eg,
> >
> > generate p p > t t~
> > output ttbar
> >
> > generate p p > w+ w-
> > output ww
> >
> > and
> >
> > generate p p > t t~ @1
> > add process p p > w+ w- @2
> > output combined_proc
> >
> > We compare the cross-sections of these processes for three different
> number of events, 10,1000 and 10000 keeping the same iseed value. However
> the cross-sections for ttbar, ww and combined_proc differ by upto 11 % for
> w+w- pair production whereas upto 0.5% for tt~ pair production in each
> case. I am unsure as to why this discrepancy occurs especially for ww where
> it is nearly 11 %.
> >
> > I was wondering whether for the combined processes, could it be that the
> number of events for computing the cross-section of the combined process
> and not the individual process is 1000 events during survey whereas for the
> processes generated separately, survey uses 1000 events and hence resulting
> in better accuracy in the results?
> >
> > Your advice in this regard will be very helpful to understand this
> issue.
> >
> > Thanking You,
> >
> > Best Regards,
> >
> > Juhi
> >
> > --
> > 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/672506/+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/672506
>
> 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.

Juhi Dutta (dutta-juhi91) said : #4

Hi,

Thank you for the reply.

>For the refine we stop when we have generated enough events if the survey generated enough then this stage is skipped

The number of events considered enough is 1000 or whatever number gives the cross-section upto 20% accuracy?

Also, is it possible to set the desired accuracy could be tuned say to 10 % or 5 % in the process or run card? here itself?

Best Regards,

Juhi

Hi,

> The number of events considered enough is 1000 or whatever number gives
> the cross-section upto 20% accuracy?

so let assume that the cross-section of
 p p > t t~ is 450 pb
and the one of p p > w+ w- is 50 pb

So if you ask for 1000 events in the run_card,
you technically ask for 950 events of t t~ and 50 events of ww (up to the square-root of such number due to the poisson distribution)
So the refine stage will stop when reaching such numbers.
At this stage the precision on the cross-section of each channel is not an issue anymore and is not used.
(Even if you have a correlation between the precision obtained for a given channel and the number of events generated in a given channel).

> Also, is it possible to set the desired accuracy could be tuned say to
> 10 % or 5 % in the process or run card? here itself?

Not via the run card. If you do not run the "generate_events" and/or "launch" command,
then you have to run first the "survey" and then the "refine" manually.
If you do that, then you can set a precision level (on the total cross-section) when running the refine command.
But this is for very advanced user only. The easiest is to know that the error is basically sqrt(N)
so from that, you can set in the run_card the number of events that you need for a given --statistical-- precision

Cheers,

Olivier

> On 23 Aug 2018, at 11:32, Juhi Dutta <email address hidden> wrote:
>
> Question #672506 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/672506
>
> Status: Answered => Open
>
> Juhi Dutta is still having a problem:
> Hi,
>
>
> Thank you for the reply.
>
>> For the refine we stop when we have generated enough events if the
> survey generated enough then this stage is skipped
>
> The number of events considered enough is 1000 or whatever number gives
> the cross-section upto 20% accuracy?
>
> Also, is it possible to set the desired accuracy could be tuned say to
> 10 % or 5 % in the process or run card? here itself?
>
> Best Regards,
>
> Juhi
>
> --
> 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 Juhi Dutta for more information if necessary.

To post a message you must log in.