How reliable are MC simulations when it comes to jets?

  • Context: Graduate 
  • Thread starter Thread starter ChrisVer
  • Start date Start date
  • Tags Tags
    Jets Simulations
Click For Summary

Discussion Overview

The discussion centers on the reliability of Monte Carlo (MC) simulations in modeling jets, particularly in the context of experimental analyses and jet calibrations. Participants explore the implications of using MC-generated jets for various applications, including tag-and-probing and training Boosted Decision Trees (BDTs).

Discussion Character

  • Debate/contested

Main Points Raised

  • Some participants express concerns about the reliability of MC simulations in modeling jet fragmentation and lepton misidentification, suggesting a preference for data-driven methods and control regions.
  • One participant questions the definition of MC, asking whether it refers to a full simulation including pileup and detector response or just a hard process with parton shower and hadronization.
  • Another participant agrees that even parton showers may not be reliable, indicating skepticism about the overall accuracy of MC simulations.
  • There is a mention that while MC is often necessary, its use can lead to significant systematic uncertainties in analyses.

Areas of Agreement / Disagreement

Participants generally express skepticism about the reliability of MC simulations, particularly regarding their application in jet modeling. However, there is no consensus on the extent of this unreliability or the implications for analyses that depend on MC-generated jets.

Contextual Notes

Participants highlight potential limitations in the accuracy of MC simulations, including dependence on the specific MC used and the associated systematic uncertainties that may arise in analyses.

ChrisVer
Science Advisor
Messages
3,372
Reaction score
465
I have been reading in papers here and there, that di/multi-jet MC generated or even other process that have one or more associated jets are used for tag-and-probing or for training several BDTs and so on. Some experiticians (theoreticians who produce experiments) are also relying on how well MCs are simulating the jets.

However I have the impression (through what I've done) that MCs are not so reliable in modelling the fragmentation or in general lepton misidentification. For that reason it's almost always the case in analyses to try and estimate those 'jets' via data-driven methods and by defining several control regions to do so.

How is that affecting people who rely a lot on jets? Or even more, people who study jet calibrations and so on as in:
https://cds.cern.ch/record/2037613/files/ATL-PHYS-PUB-2015-015.pdf
which initiated my question since they seem to take MC jets for their analysis
 
Physics news on Phys.org
Can you specify what you mean by MC?

Do you mean a full simulation of events including pileup, detector response etc.? Or do you mean just a full pp event with a hard process, Parton shower and hadronisation?

There might always be some minimal dependence on the input MC on the evaluation of the jet energy scale etc. But such dependence should be accounted for in the systematic uncertainties.
 
I guess I was referring to the whole simulation... but I think even the parton showers are not very reliable.
 
ChrisVer said:
which initiated my question since they seem to take MC jets for their analysis
You can rarely work completely without MC, but in general it is avoided as much as possible, and where it is used it often leads to large systematic uncertainties.