PDF effect on distributions

Asked by Jake on 2018-02-01

Hi,

I would like to estimate the intra-PDF effects on an invariant mass distribution. In other words, I would like to use non-central PDF values to generate events through to Delphes level.

To my understanding, running systematics incorporates the reweighting into errors in the cross-section estimate only. Is there a way to use non-central PDF values/reweighting in the simulation of the event kinematics?

This may be a stretch, but is there furthermore a way to do this that wouldn't require fully rerunning Pythia and Delphes for each reweighting?

Thanks,
Jake

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Last query:
2018-02-01
Last reply:
2018-03-09

Hi Jake,

> To my understanding, running systematics incorporates the reweighting into errors in the cross-section estimate only.

This is not correct, you have the weight for each event (and each PDF set) so you have the full distribution available.

> This may be a stretch, but is there furthermore a way to do this that wouldn't require fully rerunning Pythia and Delphes for each reweighting?

Since pythia can also depend of the PDF, you have to decide how much precise you want your generation.
(Those are actually only PDF ratio which reduce the dependency so to my point of view this is not critical)

So if you decide to neglect those effects, you can use the original weights related to the systematics.
If you want to keep such effects then you have to rerun everything.

Cheers,

Olivier

> On 1 Feb 2018, at 13:09, Jake <email address hidden> wrote:
>
> New question #663955 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/663955
>
> Hi,
>
> I would like to estimate the intra-PDF effects on an invariant mass distribution. In other words, I would like to use non-central PDF values to generate events through to Delphes level.
>
> To my understanding, running systematics incorporates the reweighting into errors in the cross-section estimate only. Is there a way to use non-central PDF values/reweighting in the simulation of the event kinematics?
>
> This may be a stretch, but is there furthermore a way to do this that wouldn't require fully rerunning Pythia and Delphes for each reweighting?
>
> Thanks,
> Jake
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Jake (hasek05-06) said : #2

Hi Olivier,

Thanks for your explanation. I have a follow up question. I ran systematics with nnpdf23lo. As far as I know there should be 101 different weights associated with this PDF set, but I see I have 145 weights per event in the unweighted event tree. Looking at the HepMC file, I notice that many of these are due to scale variations. I'm wondering about how to access these weights. For each event, is it safe to assume the first 101 weights in the leaf are simply due to the PDF, and the rest are due to scale variations?

Furthermore, I have issues with running Delphes after systematics. Delphes complains that the HepMC event format is invalid. Is this a problem with the way MG writes the file, or should I contact Delphes support?

Thanks,
Jake

Hi,

> For each event, is it safe
> to assume the first 101 weights in the leaf are simply due to the PDF,
> and the rest are due to scale variations?

Safe certainly not. The order is likely depending of the tool chain used.
I'm not an hepmc expert (since MG5aMC does not generate that format --the parton shower does--)
but If I remember correctly the leshouches convention is that they should have specific name format for such weight.

> Furthermore, I have issues with running Delphes after systematics.
> Delphes complains that the HepMC event format is invalid. Is this a
> problem with the way MG writes the file, or should I contact Delphes
> support?

Actually you should contact pythia8 support for that, I believe that this is on their side and not on Delphes side.

Cheers,

Olivier

> On 1 Mar 2018, at 19:37, Jake <email address hidden> wrote:
>
> Question #663955 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/663955
>
> Jake posted a new comment:
> Hi Olivier,
>
> Thanks for your explanation. I have a follow up question. I ran
> systematics with nnpdf23lo. As far as I know there should be 101
> different weights associated with this PDF set, but I see I have 145
> weights per event in the unweighted event tree. Looking at the HepMC
> file, I notice that many of these are due to scale variations. I'm
> wondering about how to access these weights. For each event, is it safe
> to assume the first 101 weights in the leaf are simply due to the PDF,
> and the rest are due to scale variations?
>
> Furthermore, I have issues with running Delphes after systematics.
> Delphes complains that the HepMC event format is invalid. Is this a
> problem with the way MG writes the file, or should I contact Delphes
> support?
>
> Thanks,
> Jake
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Jake (hasek05-06) said : #4

Hi Olivier,

What about in the ROOT tree generated by MadGraph? Is there a specific order to the weights in the leaf for an individual event, and is this order the same for all events?

Jake

Hi,

The Root Tree is actually generated by either Delphes and/or ExRootAnalysis.

I doubt that ExRootAnalysis support multi-wieght options so I guess that you are referring to the Delphes one.
But actually in those two case we are not the author of the tool so I can not comment.

I can only comment the details of the weight on the lhe file which is the only official output format coming out directly from our code.

Cheers,

Olivier

> On 9 Mar 2018, at 21:17, Jake <email address hidden> wrote:
>
> Question #663955 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/663955
>
> Jake posted a new comment:
> Hi Olivier,
>
> What about in the ROOT tree generated by MadGraph? Is there a specific
> order to the weights in the leaf for an individual event, and is this
> order the same for all events?
>
> Jake
>
> --
> 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 Jake for more information if necessary.

To post a message you must log in.