How to reweight scale variation

Asked by Yizhou Cai

Dear experts,

I am doing an analysis by Madgraph at LO. I already have some samples with systematics error like scale variations, and want to scan more parameter points by reweighting these samples. Is there any way to get scale variations for the new points after reweight?

In fact, I have tryed to simply compute the new weights by multiplying |ME_new|/|ME_old|, but it seems that modifying run_card.dat will not change the ME, and it must be troublesome because we have more than two hundred weights for systematics.

Regards,
Yizhou

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
Last query:
Last reply:

This question was reopened

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#1

Hi,

reweight is a very large word that can mean very different stuff.
In MG5_aMC we have two code doing rewieghing
1) the re-weight module that compute |ME_new|^2/|ME_old|^2.
In that mode the scale are taken from the events file.

2) the systematics module that compute the PDF/scale uncertainty that one does not compute the full matrix-element.
but compute those weight thanks to the meta-data present in the lhef file (so much faster)

so here I'm not sure what you mean by "after reweight". Do you mean the first one?

So assuming that it is:
you can ask the re-weighting module to re-run the systematics scale computation with the command
change systematics OPTS
where OPTS are the options for the systematics.py shell script.

Note that
- This force the output mode of the reweighting to create a new lhef file with the new weight as a new central weight when doing the re-weighting
- This assumes that all the meta-data for the re-weighting are still valid (in particular that the dependence in apha_s is not impacted)

Cheers,

Olivier

PS: more info here: https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/Reweight <https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/Reweight>
and here: https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/Systematics

> On 18 Oct 2021, at 15:35, Yizhou Cai <email address hidden> wrote:
>
> New question #699143 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/699143
>
> Dear experts,
>
> I am doing an analysis by Madgraph at LO. I already have some samples with systematics error like scale variations, and want to scan more parameter points by reweighting these samples. Is there any way to get scale variations for the new points after reweight?
>
> In fact, I have tryed to simply compute the new weights by multiplying |ME_new|/|ME_old|, but it seems that modifying run_card.dat will not change the ME, and it must be troublesome because we have more than two hundred weights for systematics.
>
> Regards,
> Yizhou
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Yizhou Cai (yzcai) said :
#2

Thanks Olivier Mattelaer, that solved my question.

Revision history for this message
Yizhou Cai (yzcai) said (last edit ):
#3

Hi Olivier,

Many thanks to your kindly reply, it is very useful information for me.

I have a further questions, since I use the MadSpin to decay, and I note that the reweight module should be done before MadSpin, could I simply use the new weights after doing this? Or I still need to use MadSpin to correct something in the output?

Regards,
Yizhou

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#4

Could you correct the typo in
"the reweight module should be done before reweighting"

Revision history for this message
Yizhou Cai (yzcai) said :
#5

Sorry, modified.

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#6

Well depend what you mean,

You can use the weight before and after madspin, they will not be the same since MadSpin will multiply all the weight by the BR.
Obviously if your BR is technically not the same for all the reweighted theory then this will be problematic.

Cheers,

Olivier

Revision history for this message
Yizhou Cai (yzcai) said (last edit ):
#7

Thanks! I think it solved my question because the particles completely decay in MadSpin.

But I am curious that, such as g g > t t~ and then decay top in MadSpin, the 4-vector of top quarks may be different before and after MadSpin. I think in this situation the matrix element will be changed, so why is the precious one be chosen? I mean, if the new weight for another param point is caculated by Weight_initial*|ME_beforeMadSpin_new|^2/|ME_beforeMadSpin_old|^2, will it be same as Weight_initial*|ME_afterMadSpin_new|^2/|ME_afterMadSpin_old|^2? Or we just trust the previous one more?

Regards,
Yizhou

Revision history for this message
Best Olivier Mattelaer (olivier-mattelaer) said :
#8

Hi,

In principle one can use reweighting before or after madspin.
Now doing it after can lead to a lot of ambiguity that the re-weighting module is not able to handle and therefore I have set it before by default.
Now nothing prevent you to run the "reweight" command on a madspin generate sample. (even if you might need to specify the syntax of the "new" process (or the various processes at NLO).

Cheers,

Olivier

Revision history for this message
Yizhou Cai (yzcai) said :
#9

Thanks Olivier Mattelaer, that solved my question.