Follow-up from "Draft: Use new MC samples in tests with data-like decoding versions"
The following discussion from !1703 (merged) should be addressed:
-
@cmarinbe started a discussion: (+10 comments)
After the changes in the input samples in !1703 (merged) although we reconstruct a similar number of clusters, the reported calo cluster efficiency is much lower. Eg from https://lhcb-nightlies.web.cern.ch/nightly/lhcb-dd4hep/1409/Moore/x86_64_v2-centos7-gcc11-opt/tests#RecoConf_hlt2_fastest_reco_without_UT_with_mcchecking:
Different content of Counters for algo CaloClusterEff_ET250
(CaloClusterEff_ET250 ref) # reconstructed | 97 | 1490 | 15.361 | 5.9503 | 1.0000 | 31.000
(CaloClusterEff_ET250 new) # reconstructed | 89 | 7 | 0.078652 | 0.26919 | 0.0000 | 1.0000
(CaloClusterEff_ET250 ref) # reconstructible | 97 | 5579 | 57.515 | 28.281 | 5.0000 | 122.00
(CaloClusterEff_ET250 new) # reconstructible | 89 | 4059 | 45.607 | 26.594 | 3.0000 | 108.00
(CaloClusterEff_ET250 ref) # signal | 97 | 26397 | 272.13 | 121.14 | 43.000 | 626.00
(CaloClusterEff_ET250 new) # signal | 89 | 38508 | 432.67 | 229.13 | 60.000 | 1052.0
(CaloClusterEff_ET250 ref) More than 1 endVertex found for signal | 95
(CaloClusterEff_ET250 new) More than 1 endVertex found for signal | 94
(CaloClusterEff_ET250 ref) reco efficiency | 5562 | 1490 | ( 26.78892 +- 0.5938141)%
(CaloClusterEff_ET250 new) reco efficiency | 4004 | 7 | (0.1748252 +- 0.06601992)%CaloClusterEfficiency might need to be adapted to the new way of creating the MC links.