Members: Andy B, Ben F, Kar D, David Y, Louie C, Sabine

Analysis combination:

• subsets of compatible SRs
• automatic determination of (approximate) orthogonality – maybe using shared samples
• Demonstration of systematic correlation between analyses from SM uncertainty breakdowns (DY, LC)
  1. LC: I mentioned yesterday the list of HEPData Entries which have uncertainty breakdowns. See here: https://gitlab.cern.ch/lhcewkwg/lhcewkwg-vjets/correlations-library/blob/master/RivetAnalysesWithErrorBreakdown.md
  2. LC: The info is now propagated into YODA files automatically when downloaded from HEPData (since Nov 2018)
  3. LC: The error breakdown can be converted to a covariance matrix with these tools quite trivially. Here are some example tools to do so from the YODA files downloaded from HEPData: https://gitlab.cern.ch/lhcewkwg/lhcewkwg-vjets/correlations-library
  4. LC: Once it is in a cov matrix form, is should be easy to use that for a simplified likelihood, AFAIK?

Full import of Rivet-HepData correlations and application to test B-L contur scan: http://www.hep.ucl.ac.uk/~yallup/LH2019Correlations.pdf

Stabilisation of SL computation:

• Generator biasing
• Automatic aggregation of SRs: Wolfgang code to cluster until some stopping criterion