Reweighting: no particle in model

I know that

'--modelname' keeps the original particle names for the model'

but I didnt manage yet to get this into the reweight_card.dat ...
If I import it there again, I get:

INFO: Running Reweight
import model Monotops_UFO --modelname
Command "generate_events " interrupted with error:
InvalidCmd : No such file or directory : model Monotops_UFO --modelname

I just checked:

Around line 1153 of the reweight_interface.py, I put a

  if use_mg_default:
            print 'Using MG default'
            base_model.pass_particles_name_in_mg_default()
        else:
            print 'NOT using MG default'

to check if the "-modelname" of the proc_card was detected. It is!

INFO: detected model: Monotops_UFO. Loading...
NOT using MG default
INFO: generating the square matrix element for reweighting

But then again I get:

INFO: generate p p > t V > b W+ psi psibar @0 ;add process p p > t~ V > b~ W- psi psibar @1 ;
Command "generate p p > t V > b W+ psi psibar @0 " interrupted in sub-command:
"generate p p > t V > b W+ psi psibar @0" with error:
InvalidCmd : No particle psi in model

Hi,

Before starting to investigate, could you confirm that you are using the lastest version of the code?
(2.4.3) A similar bug (if not the same) has been fixed in that version.

Cheers,

Olivier

Hi Olivier,

I was using 2_3_3. Indeed, moving to 2_4_3 fixed it!

Thanks,
Benedikt

Hi Olivier,

May I ask for your physical advice this time for the process generated above? I encounter large offsets in the reweighted cross sections depending on which starting point I choose.

If I calculate the cross section with a quark-dark matter coupling of g_q = 0.5, I get the following cross section:
9.21467 pb

Now I want to change the nominal quark coupling to other values and want to reproduce the xsec for g_q = 0.5 by having this point in the reweighting.

Here are the numbers I get:

nominal quark coupling: 0.7
xsef for 0.7 = 9.72541 pb
reweighting to 0.5
reweighed xsec = 5.93061227539 +- 0.0245532299163 pb

nominal quark coupling: 0.6
xsec for 0.6 = 9.60116 pb
reweighting to 0.5
reweighted xsec = 7.2059589022 +- 0.0256498850063 pb

I think the reweighting card had all the needed changes - I also modified the DECAY width of the mediator because the couplings have changed.
So I don't know why I get so crazily numbers for g_q = 0.5. Should I really assume that the shape of the matrix element has changed so much that the reweighting is not reliable anymore? I don't expect large differences in the kinematics if I change the g_q from 0.5 to 0.6 ...

Any ideas and suggestions are appreciated.

Thanks,
Benedikt

Hi

How much change your width?
If you have that mediator in s-channel and the width change a lot, then
the BW shape can explain your behavior.

The solution is typically to revert your reweighting.
This is typically fine to start from the largest width to reweight by smaller width, the opposite direction
can lead to the type of problem that you face.

Cheers,

Olivier

> On Aug 16, 2016, at 17:38, Benedikt Maier <email address hidden> wrote:
>
> Question #345458 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/345458
>
> Status: Solved => Open
>
> Benedikt Maier is still having a problem:
> Hi Olivier,
>
> May I ask for your physical advice this time for the process generated
> above? I encounter large offsets in the reweighted cross sections
> depending on which starting point I choose.
>
> If I calculate the cross section with a quark-dark matter coupling of g_q = 0.5, I get the following cross section:
> 9.21467 pb
>
> Now I want to change the nominal quark coupling to other values and want
> to reproduce the xsec for g_q = 0.5 by having this point in the
> reweighting.
>
> Here are the numbers I get:
>
> nominal quark coupling: 0.7
> xsef for 0.7 = 9.72541 pb
> reweighting to 0.5
> reweighed xsec = 5.93061227539 +- 0.0245532299163 pb
>
> nominal quark coupling: 0.6
> xsec for 0.6 = 9.60116 pb
> reweighting to 0.5
> reweighted xsec = 7.2059589022 +- 0.0256498850063 pb
>
>
> I think the reweighting card had all the needed changes - I also modified the DECAY width of the mediator because the couplings have changed.
> So I don't know why I get so crazily numbers for g_q = 0.5. Should I really assume that the shape of the matrix element has changed so much that the reweighting is not reliable anymore? I don't expect large differences in the kinematics if I change the g_q from 0.5 to 0.6 ...
>
> Any ideas and suggestions are appreciated.
>
> Thanks,
> Benedikt
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Olivier,

The width changes like this when having g_q = 0.5 instead of g_q = 0.6 :

set param_card decay 32 65.86771 # orig: 89.01738

so it is already done in this manner going from larger to smaller widths.

I quickly checked with g_q = 0.1:

set param_card decay 32 65.87567 # orig: 15.35936

xsec for 0.1 = 1.643
reweighting to 0.5
reweighted xsec = 35.9088975812 +- 0.718197836931 pb

So the problem is there in both directions.

Cheers,
Benedikt

HI,

Did you try to increase the statistics to see if the bias decrease?
Did you try to remove the helicity by helicity re-weighting? [adding change helicity none]

Otherwise, I do not know.
If you send me by email the model and all the detail to reproduce your run, I can give it a quick look to
check if I spot anything weird or if indeed the re-weighting method fails in this case.

Cheers,

Olivier

If you can send me the model and all the information to reproduce your number
> On Aug 16, 2016, at 20:17, Benedikt Maier <email address hidden> wrote:
>
> Question #345458 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/345458
>
> Status: Answered => Open
>
> Benedikt Maier is still having a problem:
> Hi Olivier,
>
>
> The width changes like this when having g_q = 0.5 instead of g_q = 0.6 :
>
> set param_card decay 32 65.86771 # orig: 89.01738
>
> so it is already done in this manner going from larger to smaller
> widths.
>
>
> I quickly checked with g_q = 0.1:
>
> set param_card decay 32 65.87567 # orig: 15.35936
>
> xsec for 0.1 = 1.643
> reweighting to 0.5
> reweighted xsec = 35.9088975812 +- 0.718197836931 pb
>
>
> So the problem is there in both directions.
>
> Cheers,
> Benedikt
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Olivier,

I tripled the statistics, but the offset remains the same.

If you could check, that would be extremely great.
I put the model here:

https://www.dropbox.com/s/cndskxgetmt7gpo/monotops_v3.ufo.tgz?dl=0

and the cards here:

https://www.dropbox.com/s/0ztxiw0xak40fo6/Cards.zip?dl=0

Let me know in case you need something else.

Cheers,
Benedikt

Dear Benedickt,

When running in debug mode. I see the following message:
> DEBUG: Command "reweight run_02 -from_cards" interrupted with error:
> DEBUG: AttributeError : 'ReweightInterface' object has no attribute 'do_compute_widths'
> DEBUG: Please report this bug on https://bugs.launchpad.net/mg5amcnlo
> DEBUG: More information is found in '/Users/omatt/Documents/eclipse/monotop_samples/fcnc_v3_reweight/run_02_tag_2_debug.log'.
> DEBUG: Please attach this file to your report.

and the reweighting stop without providing any number.
If I run in “user” mode, I do not have the printing but I have the same behavior (i.e. no value reported)
Those are clearly related to the width set on Auto which is not supported for the re-weighting.

So I’m not sure why your cards produced some result in your case but not in mine.
Anyway by hardcoding the value for the width in the reweight_card, i can reproduce your issue.

I can actually fix it if I put at the begin of the reweight_card:
change helicity False

In this mode the helicity information on each events will not be consider and the reweighting is performed by making the average over helicity.
I do not expect such radical change with the helicity precision to be set ON/OFF.
I recently changed the behavior of the reweigh tin by helicity. Looks like I need to check this mode.

Cheers,

Olivier

> On Aug 16, 2016, at 23:08, Benedikt Maier <email address hidden> wrote:
>
> Question #345458 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/345458
>
> Status: Answered => Open
>
> Benedikt Maier is still having a problem:
> Hi Olivier,
>
> I tripled the statistics, but the offset remains the same.
>
> If you could check, that would be extremely great.
> I put the model here:
>
> https://www.dropbox.com/s/cndskxgetmt7gpo/monotops_v3.ufo.tgz?dl=0
>
> and the cards here:
>
> https://www.dropbox.com/s/0ztxiw0xak40fo6/Cards.zip?dl=0
>
> Let me know in case you need something else.
>
> Cheers,
> Benedikt
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Olivier,

Thanks so much for the quick help. Indeed, with the "change helicity False" the reweighting accurately reproduces the values. I will check now kinematic distributions.

Thanks!
Benedikt

PS: I don't know either why the "Auto" for the width works for me (a user) and not for you (developer). Normally it is the other way around, right? :)

Hi Olivier,

One thing I realized is:

I previously generated the process with MG_2_3_3 and the particle 32 (the mediator) is always in the event file. After changing to MG_2_4_3 however, I only see the DM fermions of the decay in the LHE file; not a single event has a mediator.

Is this expected due to some change in the MG version?

I specified the BW cut in the run_card as very large (1000000), and anyways I wouldn't expect the BW cutoff to have an effect since I do "p p > t V > b W+ psi psibar" and not "p p > t V, V > psi psibar".

Thanks!
Benedikt

As I just discovered in this moment, it has nothing to do with 2_3_3 or 2_4_3.

It rather is related to whether or not I apply reweighting!

If I apply reweighting, the mediator has disappeared. If I do not apply it, it is still there.

Hi,

I have tested with multiple version (and indeed found that in the stable release the Auto is working for re-weighting but not in the development version) but in none of those version, I can reproduce that problem.

I have done a simple process
 p p > Z > e+ e-
and the Z is indeed present inside the events information

Note that setting bwcutoff to large value is quite bad since it will also force the intermediate particle to be systematically written in the lhe file.
Which force the shower to keep that invariant mass conserved by the shower which is not a good idea if your particle is far to be on shell.

Cheers,

Olivier

> On Aug 18, 2016, at 18:37, Benedikt Maier <email address hidden> wrote:
>
> Question #345458 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/345458
>
> Benedikt Maier posted a new comment:
> As I just discovered in this moment, it has nothing to do with 2_3_3 or
> 2_4_3.
>
> It rather is related to whether or not I apply reweighting!
>
> If I apply reweighting, the mediator has disappeared. If I do not apply
> it, it is still there.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Olivier,

Thanks for the useful information. I will set it to smaller values then.

And for what concerns the LHE file, it must somehow have to do with the model then. I also checked with p p > Z > e+ e- and confirm the Z is always there even if I apply some reweighting.

I took two screenshots corresponding to two different Event/run_xy with the monotop model.

Here the reweighting was turned on:

https://www.dropbox.com/s/ungejl2hraby9a6/Screen%20Shot%202016-08-18%20at%2015.47.34.png?dl=0

and here turned off:

https://www.dropbox.com/s/yj0b611lwmpmj1v/Screen%20Shot%202016-08-18%20at%2015.47.16.png?dl=0

As you can see, there is no particle "32" in the one with the reweighting.

But actually it is no big deal - I can reconstruct the 32 from the two dark matter particles.

Thanks again for all the help,
Benedikt

Hi,

Can you send me (either to my email or via a link) the banner of both generation?
Such that I can check if this is not linked to another change?

Cheers,

Oliver

> On Aug 18, 2016, at 20:52, Benedikt Maier <email address hidden> wrote:
>
> Question #345458 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/345458
>
> Benedikt Maier posted a new comment:
> Hi Olivier,
>
> Thanks for the useful information. I will set it to smaller values then.
>
> And for what concerns the LHE file, it must somehow have to do with the
> model then. I also checked with p p > Z > e+ e- and confirm the Z is
> always there even if I apply some reweighting.
>
>
> I took two screenshots corresponding to two different Event/run_xy with the monotop model.
>
> Here the reweighting was turned on:
>
> https://www.dropbox.com/s/ungejl2hraby9a6/Screen%20Shot%202016-08-18%20at%2015.47.34.png?dl=0
>
> and here turned off:
>
> https://www.dropbox.com/s/yj0b611lwmpmj1v/Screen%20Shot%202016-08-18%20at%2015.47.16.png?dl=0
>
>
> As you can see, there is no particle "32" in the one with the reweighting.
>
> But actually it is no big deal - I can reconstruct the 32 from the two
> dark matter particles.
>
> Thanks again for all the help,
> Benedikt
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.