lack of documentation on correlation matrices

[kratsg]

• https://hepdata-lib.readthedocs.io/en/latest/usage.html#two-dimensional-plots
• https://github.com/HEPData/hepdata_lib/blob/main/examples/correlation.ipynb [missing correlation.root file]
• https://github.com/HEPData/hepdata_lib/blob/main/examples/reading_scikithep_histograms.ipynb

In either of these two, it is not explained at all what the shape and ordering of the dependent variable is -- that is, using the description in the documentation:

table = Table()

x = Variable("Variable on the x axis",
             is_independent=True,
             is_binned=True)
# x.values = [ ... ]

y = Variable("Variable on the y axis",
             is_independent=True,
             is_binned=True)
# y.values = [ ... ]

v1 = Variable("A variable depending on x and y",
              is_independent=False,
              is_binned=False)
# v1.values = [ ... ]

does that mean v1 is a 2-dimensional value? Or a 1-dimensional flattened list? If x has M and y has N, then v1 has to have up to M*N values -- but how do you actually describe the matrix? I don't see this described anywhere in the documentation.

[clelange]

Thanks for the input. We can try to improve the docs over the next couple of weeks.

[GraemeWatt]

See the main submission documentation on Correlation/covariance matrices. The correlation.ipynb notebook writes the correlation.root file in the first step, so it is not included in the repository. The notebook calls read_hist_2d which in turn calls get_hist_2d_points. If the original correlation matrix has M*N values, then x, y and v1 are all 1-dimensional lists with M*N values. The ordering is not important. See also the Sandbox record linked at the bottom of the correlation.ipynb notebook.

I agree that the hepdata_lib docs could be improved, perhaps by linking to the relevant section of the main submission documentation. Feel free to submit a PR!