Closed ClarkMcGrew closed 1 month ago
Strike the last comment because I didn't notice differences between LTS and main. The LTS sees agreement between the MIGRAD results and the HESSE results (both values and errors, but the values "must" be the same). The main code is converging to a nearby, but different point, and is getting reporting different values for MIGRAD and HESSE.
This branch is ready to be merged into LTS/1.8.x.
The "reconstructed" histograms were being constructed with one of the bins truncated. That caused some of the events to be missing in the likelihood. This fixes the histograms to cover the entire region, and then updates the expected test values for the new binning. It also sets an explicit tolerance in the config file so that the results are more stable.
Note: Setting the tolerance to
1.0E-6
fixes a problem noted during a working session that the MIGRAD errors were very inaccurate. The MIGRAD errors now agree rather well with the HESSE errors.