Line scan verification – prof-scanchi2¶To verify that an optimal set of parameter values, that was found by running prof-tune and inspecting the scatter of the results (kebab plots) with prof-plotresultscatter, the Professor tool chain provides prof-scanchi2. At certain points in parameter space, prof-scanchi2 compares the goodness of fit (GoF) between the “true” MC data and the reference data with the GoF between the interpolation and the reference data. The latter is the GoF that was used in the optimization procedure in prof-tune, the first is the GoF that is really to be minimized. The points are usually sampled along a straight line with prof-sampleparams using the line scan mode. Depending on the given arguments prof-scanchi2 plots only the “true” MC-reference data GoF or the interpolation-reference data GoF. The --datadir DATADIR and related options are used as normal to specify the reference data, MC runs, and interpolation objects: see the path options page. To plot the “true” MC-reference data GoF the location of the scan MC data must be specified using --scandir SCANDIR. By default DATADIR/scandir/ is assumed (see --datadir below). To plot the interpolation-reference data GoF the location of the interpolation files and a list of run combinations must be given. The location of the interpolation files can be passed by --ipoldir IPOLDIR (default: DATADIR/ipoldir/). Run combinations can be passed either by --runcombs RUNSFILE:IPOLMETHOD or they are taken from minimization results files passed by --results RESULTS. Note If no MC scan directory exists, prof-scanchi2 will fail unless --no-scandata is specified and the end points of the line are passed by --endpoints ENDPOINTS. By default the x-axis represents the location along the line. Using --params PARAM1[,...] the x-axis will represent the projected position on the respective parameter axis. Interpretation¶A good agreement between the two GoF definitions shows that the initial assumption to approximate the per-bin response function by the interpolation is valid. If the comparison shows a difference, it is advisable to resize the initial parameter space (i.e. in- or excluding some regions of parameter space) and repeat the full procedure of MC sampling, interpolating and GoF optimization. If time is critical, the extract mode of prof-runcombs can be used to limit the used MC sample data to narrower parameter ranges. Examples¶Plot the “true” MC-reference GoF of the MC runs in my-scan-data/. The end points of the line are taken from the used_params files and the observables/weights for the GoF are taken from my-weights: prof-scanchi2 --scandir my-scan-data --weights my-weights
Plot the interpolation-reference GoF along a straight line between the two points given by endpoints.params (no MC scan data is plotted). The interpolations are taken from my-ipol/, the observables/weights are taken from my-weights, and the run combinations are taken from my-runs. Quadratic polynomials are used for interpolation: prof-scanchi2 --ipoldir my-ipol --weights my-weights --runcombs my-runs:quadratic --no-scandata --endpoints endpoints.params
Combine the above two examples and directly compare “true” MC-reference GoF and interpolation-reference GoF: prof-scanchi2 --scandir my-scan-data --weights my-weights --ipoldir my-ipol --runcombs my-runs:quadratic
Command-line options¶The --datadir DATADIR and related options are used as normal to specify the reference data, MC runs, and interpolation objects: see the path options page.
|