Model-Specific Closure Test

Colibri can be used to run model-specific closure tests. Here we show how to do so for the Les Houches parametrisation (implemented here).

Step 1: runcard

The runcard runcards/lh_closure_test_model.yaml in the examples/les_houches_example directory runs such a closure test. It is the same as the runcard presented here, except that closure_test_pdf: LH_PARAM_20250519 is replaced by the following block:

closure_test_pdf: colibri_model        # Generate the pseudodata with a colibri model.
closure_test_model_settings:
  model: les_houches_example           # Name of the installed colibri model package.
  parameters:                          # One value per model parameter.
    alpha_gluon: 0.356
    beta_gluon: 10.9
    alpha_up: 0.718
    beta_up: 3.81
    epsilon_up: -1.56
    gamma_up: 3.30
    alpha_down: 1.71
    beta_down: 10.0
    epsilon_down: -3.83
    gamma_down: 4.64
    norm_sigma: 0.211
    alpha_sigma: -0.048
    beta_sigma: 2.20

Within the closure_test_model_settings, you need to specify any of the hyperparameters relevant to build the model instance. The example model shown here doesn’t need further settings, but other models may need additional hyperparameters to tune.

Note that, unlike for a standard closure test, in this case we need to specify the initial values for each parameter of the PDF model. For the values above, we have taken the best-fit values, taken from Ref. A+05.

Step 2: running the fit

You can produce the fit by running the following command from the examples/les_houches_example directory:

les_houches_exe runcards/lh_closure_test_model.yaml

Running the fit will create a directory called lh_closure_test_model, where you will find the output of the fit. You can read more about it in this tutorial.

You can evolve it by following the instructions given in this tutorial.