C2 2D scan best-fit value depends on the POI order
Hi @mrieger,
I was running the C_2 scan using the associate physics model (still blinded), and noticed the 2D likelihood scan returned as best fit values \kappa_{\lambda}=0 and C_2=1.
The command I used is
law run PlotLikelihoodScan --version bbww_FR2 --datacards '/afs/cern.ch/work/f/fbury/datacards/UCL/bbww/NonResonant/SM/datacard_fit_NonResonant_syst_FR2_C2_v29/*txt' --pois kl,C2 --save-ranges --hh-model dhi.models.hh_model_C2klkt:model_default --scan-parameters C2,-5,5,51:kl,-25,25,51 --LikelihoodScan-workflow htcondor --LikelihoodScan-tasks-per-job 1 --LikelihoodScan-max-runtime 6hAfter discussing with Torben I tested switching the POIs and scans (--pois C2,kl) and it produced the appropriate result, this seems like a bug of the plotting script.
Additionally, as you can see some pixels have very small values. I checked the values returned by the fits, they were fine, and it turns out that the default interpolation method (root, using TGraph2D) is the culprit. I have ran with --interpolation-method linear and it now seems fine. I guess on that note it's just the ROOT method not working properly and there is not much to correct (although it is weird that it happens in one order of POIs and not the other). Maybe a warning somewhere that other methods can resolved that.
Florian