As soon as a minimization was successful one will immediately be interested in how well the corresponding set of parameters is doing compared to other tunings and reference data.
One could (should) in principle always produce a new Monte Carlo sample using this parameter setting, which might be okay (in terms of time consumption) if the jobs a re not too big (e.g. for LEP multiplicity ratios).
If, on the other hand, the production is more of a technical pain, e.g. involves merging histograms etc., it might be more useful to investigate what the interpolation itself predicts at the newly found minimum.
Therefore, whenever a minimization was successful, a histogram is being written to a file in flat format that holds the bin contents as calculated from the bins parameterization itself at the predicted minimum.
how to do it
In Rivet there are three tools that you will have to use one after another. First of all you have to convert the histogram file into an aida file format:
Which will create a file ipolhistos.aida in the same directory.
Then you have to use compare-histos to produce comparison plots, e.g. to reference data and maybe one (ore more) tuning run(s), say S0.aida:
compare-histos -o compareplots ../ref/*.aida ipolhistos.aida:"Professor" S0.aida:"tune S0"
This will produce one file (flat format .dat) for each observable that is in at least 2 of the files to compare. The output is written to the directory specified via -o in this example compareplots.
If you then change into the directory compareplots you can simply do
which will produce very nice ps files like in this example:
Optionally, you could use makegallery.py from the professor-package to produce an html-gallery of all the ps-files in compareplots.
In order to do so, you would have to run
src/professor/trunk/professor/tools/makegallery.py -s compareplots ps compareplots.html
That's about it!