Syntax clarifications for systematics program

Asked by Kenneth Long on 2017-07-31

Hi experts,

I'm working on transitioning to the new "systematics" module for PDF and scale reweighting, and could use a few clarifications. I'm working in version 2.5.5.

1) At NLO, just specifying 'True = store_rwgt_info' in the run_card doesn't seem to be enough for the reweight info to be stored/usable. It seems to also require 'True = reweight_scale' or 'True = reweight_PDF' to be specified.

2) Since I want to include scale weights anyway, (1) isn't really a problem. But when I try to reweight after having specified both 'True = reweight_scale,' and 'True = store_rwgt_info,' I can't figure out how to avoid having the scale weights rewritten. I tried both

(a) systematics run_01 --pdf=XXX,YYY --dyn=-1
(b) systematics run_01 --pdf=XXX,YYY --dyn=-1 --together=mur,muf

but in both cases I end up with a reweight block as copied below. Is there a syntax for this that I'm missing?

3) Is it possible to specify the ID (or starting number) of the weight? The way the previous on the fly NLO weighting switched from 1XXX to 2XXX for scale weights to PDF weight was kind of useful if the number of scale weights being stored is variable.

Thanks!

Kenneth

<weightgroup name='scale_variation -1' combine='envelope'>
      <weight id='1001'> dyn= -1 muR=0.10000E+01 muF=0.10000E+01 </weight>
      <weight id='1002'> dyn= -1 muR=0.20000E+01 muF=0.10000E+01 </weight>
      <weight id='1003'> dyn= -1 muR=0.50000E+00 muF=0.10000E+01 </weight>
      <weight id='1004'> dyn= -1 muR=0.10000E+01 muF=0.20000E+01 </weight>
      <weight id='1005'> dyn= -1 muR=0.20000E+01 muF=0.20000E+01 </weight>
      <weight id='1006'> dyn= -1 muR=0.50000E+00 muF=0.20000E+01 </weight>
      <weight id='1007'> dyn= -1 muR=0.10000E+01 muF=0.50000E+00 </weight>
      <weight id='1008'> dyn= -1 muR=0.20000E+01 muF=0.50000E+00 </weight>
      <weight id='1009'> dyn= -1 muR=0.50000E+00 muF=0.50000E+00 </weight>
    </weightgroup>
<weight id="1010" MUR="1.0" MUF="1.0" PDF="292200" > </weight>
<weightgroup name="Central scale variation" combine="envelope">
<weight id="1011" MUR="0.5" MUF="0.5" PDF="292200" > MUR=0.5 MUF=0.5 </weight>
<weight id="1012" MUR="0.5" MUF="1.0" PDF="292200" > MUR=0.5 </weight>
<weight id="1013" MUR="0.5" MUF="2.0" PDF="292200" > MUR=0.5 MUF=2.0 </weight>
<weight id="1014" MUR="1.0" MUF="0.5" PDF="292200" > MUF=0.5 </weight>
<weight id="1015" MUR="1.0" MUF="2.0" PDF="292200" > MUF=2.0 </weight>
<weight id="1016" MUR="2.0" MUF="0.5" PDF="292200" > MUR=2.0 MUF=0.5 </weight>
<weight id="1017" MUR="2.0" MUF="1.0" PDF="292200" > MUR=2.0 </weight>
<weight id="1018" MUR="2.0" MUF="2.0" PDF="292200" > MUR=2.0 MUF=2.0 </weight>
</weightgroup> # scale

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
Rikkert Frederix Edit question
Solved by:
Kenneth Long
Solved:
2017-08-02
Last query:
2017-08-02
Last reply:
2017-08-02

Hi Kenneth,

> 1) At NLO, just specifying 'True = store_rwgt_info' in the run_card doesn't seem to be enough for the reweight info to be stored/usable. It seems to also require 'True = reweight_scale' or 'True = reweight_PDF' to be specified.

Correct, you can ask to have only '1.0' for the variation of the scales and therefore limit that part if needed.

> 2) Since I want to include scale weights anyway, (1) isn't really a problem. But when I try to reweight after having specified both 'True = reweight_scale,' and 'True = store_rwgt_info,' I can't figure out how to avoid having the scale weights rewritten. I tried both
>
> (a) systematics run_01 --pdf=XXX,YYY --dyn=-1
> (b) systematics run_01 --pdf=XXX,YYY --dyn=-1 --together=mur,muf
>
> but in both cases I end up with a reweight block as copied below. Is there a syntax for this that I'm missing?

Indeed, I only add weights so far, but I will add in 2.5.6 a syntax to remove some/all weights.
Something like
--remove_wgts=all
--remove_wgts=min,max
--remove_wgts=value
--keep_wgts=min,max #keep_wgts having precedence over remove_wgts
--keep_wgts=value

> 3) Is it possible to specify the ID (or starting number) of the weight? The way the previous on the fly NLO weighting switched from 1XXX to 2XXX for scale weights to PDF weight was kind of useful if the number of scale weights being stored is variable.

I do not have such options. But this can do that for 2.5.6
Something like
--start_id=XXXXX

Anything else?

Cheers,

Olivier

> On 31 Jul 2017, at 23:09, Kenneth Long <email address hidden> wrote:
>
> New question #654417 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/654417
>
> Hi experts,
>
> I'm working on transitioning to the new "systematics" module for PDF and scale reweighting, and could use a few clarifications. I'm working in version 2.5.5.
>
> 1) At NLO, just specifying 'True = store_rwgt_info' in the run_card doesn't seem to be enough for the reweight info to be stored/usable. It seems to also require 'True = reweight_scale' or 'True = reweight_PDF' to be specified.
>
> 2) Since I want to include scale weights anyway, (1) isn't really a problem. But when I try to reweight after having specified both 'True = reweight_scale,' and 'True = store_rwgt_info,' I can't figure out how to avoid having the scale weights rewritten. I tried both
>
> (a) systematics run_01 --pdf=XXX,YYY --dyn=-1
> (b) systematics run_01 --pdf=XXX,YYY --dyn=-1 --together=mur,muf
>
> but in both cases I end up with a reweight block as copied below. Is there a syntax for this that I'm missing?
>
> 3) Is it possible to specify the ID (or starting number) of the weight? The way the previous on the fly NLO weighting switched from 1XXX to 2XXX for scale weights to PDF weight was kind of useful if the number of scale weights being stored is variable.
>
> Thanks!
>
> Kenneth
>
> <weightgroup name='scale_variation -1' combine='envelope'>
> <weight id='1001'> dyn= -1 muR=0.10000E+01 muF=0.10000E+01 </weight>
> <weight id='1002'> dyn= -1 muR=0.20000E+01 muF=0.10000E+01 </weight>
> <weight id='1003'> dyn= -1 muR=0.50000E+00 muF=0.10000E+01 </weight>
> <weight id='1004'> dyn= -1 muR=0.10000E+01 muF=0.20000E+01 </weight>
> <weight id='1005'> dyn= -1 muR=0.20000E+01 muF=0.20000E+01 </weight>
> <weight id='1006'> dyn= -1 muR=0.50000E+00 muF=0.20000E+01 </weight>
> <weight id='1007'> dyn= -1 muR=0.10000E+01 muF=0.50000E+00 </weight>
> <weight id='1008'> dyn= -1 muR=0.20000E+01 muF=0.50000E+00 </weight>
> <weight id='1009'> dyn= -1 muR=0.50000E+00 muF=0.50000E+00 </weight>
> </weightgroup>
> <weight id="1010" MUR="1.0" MUF="1.0" PDF="292200" > </weight>
> <weightgroup name="Central scale variation" combine="envelope">
> <weight id="1011" MUR="0.5" MUF="0.5" PDF="292200" > MUR=0.5 MUF=0.5 </weight>
> <weight id="1012" MUR="0.5" MUF="1.0" PDF="292200" > MUR=0.5 </weight>
> <weight id="1013" MUR="0.5" MUF="2.0" PDF="292200" > MUR=0.5 MUF=2.0 </weight>
> <weight id="1014" MUR="1.0" MUF="0.5" PDF="292200" > MUF=0.5 </weight>
> <weight id="1015" MUR="1.0" MUF="2.0" PDF="292200" > MUF=2.0 </weight>
> <weight id="1016" MUR="2.0" MUF="0.5" PDF="292200" > MUR=2.0 MUF=0.5 </weight>
> <weight id="1017" MUR="2.0" MUF="1.0" PDF="292200" > MUR=2.0 </weight>
> <weight id="1018" MUR="2.0" MUF="2.0" PDF="292200" > MUR=2.0 MUF=2.0 </weight>
> </weightgroup> # scale
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Kenneth Long (kdlong-e) said : #2

Thanks Olivier,

I still end up with 1 redundant weight for uR = uF = u_ref in the first case, correct? This isn't such a big deal, but for consistency with samples produced with older versions of MadGraph, not having any duplication would be nice. I guess there is no way to exclude this variation in the reweighting step?

The two features you mention would both be appreciated. If you don't mind sharing a patch with us when you have it, that'd also be great.

One further question: do the dynamic scale weights actually make sense in the FxFx case (or even MLM) or do you need to use the default choice for consistency in the merging?

Thanks again!

Kenneth

Hi,

> I still end up with 1 redundant weight for uR = uF = u_ref in the first case, correct?

That's right.

> The two features you mention would both be appreciated. If you don't mind sharing a patch with us when you have it, that'd also be great.

Send you the patch very soon. I have also added the option for regular expression.
i.e. --remove_weights=\d*
which will keep all the weight from thje re-weighting module but will remove all the default one of the built-in systematics module.

> One further question: do the dynamic scale weights actually make sense in the FxFx case (or even MLM) or do you need to use the default choice for consistency in the merging?

This is a complex question:

For MLM,this is done in a fully consistent way. Meaning that only the couplings associated to the renormalisation scale are going to be modifed according to the dynamical scale choice. Since QCD radiation are not associated to the renormalisation scale anymore, this variation will be smaller that you can naively expect.
Note that for MLM, you have the variation of the "alps" parameter which allowed to change the scale associated to those QCD radiation (not the functional form, only via a numerical factor) which should compensate for the smaller dependence in mur.

For FxFx, I'm not sure of the situation, and I will let Rikkert to anser this.

Cheers,

Olivier

Rikkert Frederix (frederix) said : #4

Dear Kenneth,

For FxFx only the dynamical_scale should be used for consistency in the merging.

Best,
Rikkert

Rikkert Frederix (frederix) said : #5

*default dynamical_scale

Kenneth Long (kdlong-e) said : #6

Thanks Olivier and Rikkert!