SysCalc fails at Pythia level

Hi,

could it be that my problems are due to SysCalc not being properly installed?

Calling "install SysCalc" form MadGraph fails because the lhapdf-config is not in my PATH. If I add the lhadpf-config to the PATH, the installation works, but SysCalc still fails because it does not find the shared libraries. I fixed that by adding "-Wl,-rpath $(shell lhapdf-config --libdir)" to the LIBS variable in SysCalc/src/Makefile and running make in SysCalc, but I still have the problem described in my question above.

Is there anything else the "install SysCalc" command does besides downloading and untarring the source and running make?

Am I right that, in the end, MadGraph should print the systematic error in percent on the screen?

Best,

Eric

Hi Eric

I would suggest before trying to install SysCalc , make sure that you have properly defined the env. variables like (assuming you install LHAPDF in /local dir

## Set environment variables
export PATH=$PWD/local/bin:$PATH
export LD_LIBRARY_PATH=$PWD/local/lib:$LD_LIBRARY_PATH
export PYTHONPATH=$PWD/local/lib64/python2.6/site-packages:$PYTHONPATH

Apart from that , the SysCalc output you quote in your first post looks ok - And indeed, SysCalc only prints the new cross-sections ie like you have

cale fact cross-section :8 10764.5
alpha_s emission scale fact cross-section :0 9473.98
alpha_s emission scale fact cross-section :1 9388
alpha_s emission scale fact cross-section :2 9319.83
pdf reweighted cross-section :0 9388
...
pdf reweighted cross-section :52 9300.45
min/max for the set:9174.34 9624.49
nb of event for qcut :0 30 8934
nb of event for qcut :1 50 8899

Hi Eric,

For all the SysCalc specific question, I will let Alexis to answer those.
I would however answer those related to MG5aMC

> Is there anything else the "install SysCalc" command does besides
> downloading and untiring the source and running make?

Not that much more at least (t is trying to link to the lhapdf version specified to MG5aMC code but behind that nothing)

> Am I right that, in the end, MadGraph should print the systematic error
> in percent on the screen?

No, MG5aMC do not try to read/parse the result of SysCalc.
We just run it and put a link to the final log/results in the html page.
(the lhe files is also updated with the additional weight)

Cheers,

Olivier

> On May 13, 2016, at 14:02, Eric <email address hidden> wrote:
>
> Question #292500 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/292500
>
> Eric gave more information on the question:
> Hi,
>
> could it be that my problems are due to SysCalc not being properly
> installed?
>
> Calling "install SysCalc" form MadGraph fails because the lhapdf-config
> is not in my PATH. If I add the lhadpf-config to the PATH, the
> installation works, but SysCalc still fails because it does not find the
> shared libraries. I fixed that by adding "-Wl,-rpath $(shell lhapdf-
> config --libdir)" to the LIBS variable in SysCalc/src/Makefile and
> running make in SysCalc, but I still have the problem described in my
> question above.
>
> Is there anything else the "install SysCalc" command does besides
> downloading and untarring the source and running make?
>
> Am I right that, in the end, MadGraph should print the systematic error
> in percent on the screen?
>
> Best,
>
> Eric
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Alexis and Olivier,

I reinstalled SysCalc after setting the environment variables mentioned in Alexis' answer, but MadGraph still tells me that SysCalc fails when calculating the systematics for pythia level.

Could it be that this error is not caused by SysCalc but by pythia-pgs? Apparently, pythia does not find the pdf set I use. In the tag_1_pythia.log it says:

 ==== PYTHIA WILL USE LHAPDF ====
 WRONG LHAPDF set number = 11000 given! STOP EXE!
 WILL USE 10041 INSTEAD

where 10041 is the LHAID set in the pythia card. I downloaded the CT10nlo.LHgrid to pythia-pgs/src/PDFsets, but it still does not work. Do you know if there's a way to make pythia 6 work with my LHAPDF 6 installation?

Regarding the output of SysCalc, if I compute cross sections in NLO mode, I get PDF and scale uncertainties in per cent. Are these calculated based on something similar to the output I get from SysCalc?

Cheers,

Eric

Hi Eric,

> Do
> you know if there's a way to make pythia 6 work with my LHAPDF 6
> installation?

pythia-pgs work only with LHAPDF4 (which is provide with the package)

> Regarding the output of SysCalc, if I compute cross sections in NLO
> mode, I get PDF and scale uncertainties in per cent. Are these
> calculated based on something similar to the output I get from SysCalc?

Yes.

Cheers,

Olivier

> On May 17, 2016, at 14:02, Eric <email address hidden> wrote:
>
> Question #292500 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/292500
>
> Eric posted a new comment:
> Hi Alexis and Olivier,
>
> I reinstalled SysCalc after setting the environment variables mentioned
> in Alexis' answer, but MadGraph still tells me that SysCalc fails when
> calculating the systematics for pythia level.
>
> Could it be that this error is not caused by SysCalc but by pythia-pgs?
> Apparently, pythia does not find the pdf set I use. In the
> tag_1_pythia.log it says:
>
> ==== PYTHIA WILL USE LHAPDF ====
> WRONG LHAPDF set number = 11000 given! STOP EXE!
> WILL USE 10041 INSTEAD
>
> where 10041 is the LHAID set in the pythia card. I downloaded the
> CT10nlo.LHgrid to pythia-pgs/src/PDFsets, but it still does not work. Do
> you know if there's a way to make pythia 6 work with my LHAPDF 6
> installation?
>
> Regarding the output of SysCalc, if I compute cross sections in NLO
> mode, I get PDF and scale uncertainties in per cent. Are these
> calculated based on something similar to the output I get from SysCalc?
>
> Cheers,
>
> Eric
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi,

Can you provide more information on how those uncertainties are calculated at NLO? Do they just take the relative difference between the cross sections calculated with scale factor 2 (0.5) and scale factor 1 as scale uncertainty, and the RMS of the pdfs as pdf uncertainty?

Best,

Eric

Thanks Olivier Mattelaer, that solved my question.