build for going solo

[WolfgangWaltenberger]

Hey guys, I am trying to build ColliderBit/examples/solo.cpp, I have successfully built gambit proper.
I am somehwat lost as to finding all the right -I, -l and -L flags. I did not find any help on that in the documentation, either.
Would it be possible for one of guys to help me out here? I assume it's on purpose that there aren't any build files in the examples directories?

[gambit 2.4.4] Thanks in advance

Wolfgang

[WolfgangWaltenberger]

Ah never mind, found the target in the toplevel makefile. sorry for the bother.
Wolfgang

[WolfgangWaltenberger]

Actually, sorry for the noise, let me reopen this. I cant find a simple recipe to find the solo executable for the colliderbit. Could someone please gently point me in the right direction here?

cheers

Wolfgang

[anderkve]

Hi @WolfgangWaltenberger!

As mentioned via email, ColliderBit Solo isn't 100% polished or documented yet, but it should be working. The build target is "CBS".

I just now tried to build CBS from a fresh clone of gambit 2.4.4 on my laptop with the following commands:

mkdir build ; cd build

cmake -DEIGEN3_INCLUDE_DIR="/path/to/my/eigen-3.4.0" -DWITH_YODA=True -DWITH_HEPMC=True -DWITH_ROOT=True -DCMAKE_CXX_COMPILER=g++-11 -DCMAKE_C_COMPILER=gcc-11 -DCMAKE_Fortran_COMPILER=gfortran-11 ..

make CBS

That seems to have worked as expected. So try something along those lines and let me know how it goes.

[WolfgangWaltenberger]

Perfect, that worked! The only think I was missing was the name of the target, CBS. I have been trying targets
that had the word "solo" in it. I can run it, so now I need a yaml file that is configured for CBS to take hepmc as input,
and spit out signal efficiencies (or signal yields) as output. Can I please ask you to point me to a decent starting point
of yaml file for this? Much appreciated! Wolfgang

[anderkve]

There's an example yaml file in ColliderBit/examples/solo_example.yaml -- see here: https://github.com/GambitBSM/gambit_2.4/blob/release_2.4/ColliderBit/examples/solo_example.yaml

So try running ./CBS ColliderBit/examples/solo_example.yaml.

Note that this is a minimal example -- there are actually many more options that can be added in the yaml file. I can send you a more complete yaml file soon.

One more thing: For CBS to work you probably also need to build two more backend codes for GAMBIT. So do the following before running CBS:

cd build
make nulike
make ATLAS_FullLikes

(If you forget, you should get a sensible error message when CBS starts.)

[WolfgangWaltenberger]

Halleluiah, it is working:

...
Signal region SRB bveto [150, 230] (SR index 12):
Observed events: 18
SM prediction: 15.7 +/- 2.3
Signal prediction: 167.073 +/- 167.073
Log-likelihood: -0.366338
...

I am so exuberantly happy right now ;) One quick question, the output above was a result of 820 hepmc events.
I would have thought that like the other numbers in this list, the signal prediction is given as yields, thus integer numbers.
Why am I seeing real-valued numbers?
Thanks a ton!

Wolfgang

[andrewfowlie]

167.073 +/- 167.073 looks strange as well.

[WolfgangWaltenberger]

It seems to correspond to 1 event. I wanted to create 1000 events, 820 survived the generators' cuts, I sent them through CMS-SUS-20-001/CMS_13TeV_2OSLEP_137invf. That analysis has an integrated lumi of 137/fb.
The summary cutflow table reads:

0 All events 820
1 Baseline preliminary selections 32
2 Baseline lepton selections 7
3 Signal regions selections 1

Thats +/- all i know for now.

Wolfgang

[anderkve]

Yes, the printed signal prediction is scaled to the cross-section * luminosity, hence the non-integer value. The luminosity is hard-coded in the given analysis source file, and you set the total production cross-section for your parameter point in the CBS yaml file.

[WolfgangWaltenberger]

right silly me, i could have guessed. sorry for the noise
Wolfgang

[WolfgangWaltenberger]

Success! So below, you see the first two SModelS validation plots produced from colliderbit "efficiency maps".

So:

*) a first version of the machinery is technically working! Yay!
*) numerically something is still off (the black lines and the blue lines would have to approximately coincide)
*) as these are my very first such plots, the remaining numerical problem is most likely with me
*) we will of course track the usage of colliderbit/gambit in our results' meta data, and will cite colliderbit, whenever
a colliderbit-produced result gets used (and tell our users to do so as well).
*) just to be sure the colliderbit implementation of an analysis is ok, do you guys collect some kind of validation material?
*) The last point will be particularly important for us once our validation results are "kinda/almost good". We will want to
know if the "kinda/almost" is due to us or due to colliderbit.
*) JFYI, these first plots are CMS-SUS-20-001 == CMS_13TeV_2OSLEP_137invfb. I strongly assume yes, but just to be sure,
from the gambit side this analysis should be good, right?
*) our signal efficiencies (what ATLAS calls "A x epsilon") are computed for a "cross_section_pb: 0.001",
as: eff := "Signal prediction" / (lumi[fb]).

Thanks a ton,

Wolfgang

[WolfgangWaltenberger]

ah wait, you said n := Y * lumi, okay, maybe the / lumi is wrong. will try.

[WolfgangWaltenberger]

ah wait, you said n := Y * lumi, okay, maybe the / lumi is wrong. will try.

nah, that's not it. that's correct how i did it.

[WolfgangWaltenberger]

I think I understand. You guys could only harvest about half of the signal regions, mostly the ones for hadronic signatures. Thats why for gluino gluino production, gluino -> qq N2, N2 -> Z N1, things tend to work. Not for the ewkino case -- no SR survived. For the sleptons case you guys got half the SRs to work.