Code structure

In the following the provide a quick overview of how the code works. We refer to the tutorial for a step by step guide on how to use it.

The information required to set up the analysis are read from an input runcard. The code will load the specified datasets, collecting the full experimental information and the Standard Model and EFT corrections to build the corresponding theory predictions. In the same input card the operators whose Wilson coefficients should be fitted are specified. If some of them do not appear in any of the selected datasets, the code will stop with an error message.
Using the loaded information the loglikelihood is build as a function of the Wilson coefficients to fit. This is then given as input to an Optimizer object, encoding all the specifications of the minimizer used in the analysis. The provided options are Nested Sampling and MonteCarlo, based on the external Utranest library and on an internal implementation respectively. In the first case the output is the full posterior distribution for the fitted Wilson coefficients, while in the second is the result for a single MC replica. To get the full posterior using MC, after producing a sufficient number of replicas the code must be run in the PostFit mode, which discards replicas having chi^2 higher than a certain threshold set by the user and produce the final output file.
The final output of the fit is a json file having the following structure
```
{
  O_1 : [sample_1, sample_2, ...],
  O_2 : [sample_1, sample_2, ...],
  ...
}
```
where each entry correspond to a list of samples of the Wilson coefficient corresponding to the specified operator. The final posterior is therefore provided as a discrete set of samples for each fitted Wilson coefficient.
The code can be run in the Report mode in order to produce a report to visualize the results. The different functionalities currently supported in the report are documented here