Histograms miss information to correctly compute some statistical values
Right now the histograms (Gaudi::Accumulators::Histogram
) only keep the bins and counts in each bin but without saving extra information some statistical quantities can't be computed accurately and depend on the choice of bins (even for unweighted histograms). For example, to compute the standard deviation (that will typically appear in plots done after exporting to ROOT) one would need to keep the sum of the weights and the sum of the square of the weights. I think the profile histogram (Gaudi::Accumulators::ProfileHistogram
) has the same problem and in principle the information could be recovered by iterating over every bin to get the sum of the weights and the sum of the square of the weights for every bin and them adding them up, but it doesn't trivial and I imagine most people (myself included) expected that it would be easy to get these statistical quantities. Are there any plans to change this?