Combine v11, ready for CERN SWAN
A pinned, isolated notebook environment for running CMS Combine without replacing SWAN's own Python stack.
The short path
Download the project from Git in SWAN:
Then open a terminal in the checkout and run:
The installer creates the complete CMSSW area at:
Refresh JupyterLab, open the supplied notebook, and select:
For later sessions, configure the repository's swan_env.sh as the SWAN Environment script. The setup guide walks through this once, including the kernel-picker restart behavior.
What is isolated
The parent SWAN notebook server keeps its selected LCG Python and libraries. Only the dedicated Combine kernel activates CMSSW. This boundary prevents the two Python stacks from contaminating each other while keeping both available in the same SWAN project.
flowchart LR
A[SWAN server<br/>LCG Python] --> B[combine-v11<br/>kernel launcher]
B --> C[CMSSW_16_0_0<br/>PyROOT + ipykernel]
C --> D[Combine v11.0.0]
Verified on SWAN
Reference execution: 1 August 2026. The committed notebook imports PyROOT and HiggsAnalysis.CombinedLimit, checks the exact Combine tag, exercises the main command-line tools, and completes a blind AsymptoticLimits run that produces a non-empty ROOT file.
That is an installation and runtime validation. It does not independently validate a user's physics model, datacard, nuisance model, or statistical interpretation. See Validation for the exact boundary.