First bin incorrect in HistFactory workspaces

[beojan]

When I convert a multi-channel RooStats::HistFactory::Measurement to a workspace using MakeModelAndMeasurementsFast or MakeCombinedModel, the first bin of the data histograms in each channel is much larger than it should be.

While investigating, I converted a single channel using MakeSingleChannelModel. Here, the first bin has the correct value. (The bin errors are now equal to the bin values, but I think this is because the binned data is saved as an unbinned dataset with one entry per bin, with the weight set equal to the bin value. This seems to be corrected in the combined model though.)

[bellenot]

I think @StephanH can most probably help you

[StephanH]

The explanation is here:

Extracting histograms from HistFactory generated RooWorkspace

I figured out what was happening. I didn’t cut on channelCat, and because each channel has it’s own observable, when one observable is plotted, the events from the other channels all end up in one bin. I’m not sure why this one bin isn’t an over / underflow bin. root [4] d->get(0)->Print("V"); d->weight(); 1) RooCategory:: channelCat = resolved_4b_2015(idx = 0) 2) RooRealVar:: obs_x_resolved_4b_2015 = 266.667 3) RooRealVar:: obs_x_resolved_4b_2016 = 266.667 4) RooRealVar:: …