Weight Parameters

Python Version License

Defines the configuration for evaluating the model weights in the analysis.

weight_calcula_configuration:
    number_weight_evaluation_per_chain: 20
    number_processors_used_weights: 20
    stability_factor: 1e-8
    pareto_smoothing: True

Parameters

  • number_weight_evaluation_per_chain: int

    Number of weight evaluations to perform per MCMC chain. A suggested value would be equal or larger than 20.

  • number_processors_used_weights: int

    Number of processors used in parallel for weight computations. This value has to be lower or equal to the value set in number_weight_evaluation_per_chain.

  • stability_factor: float

    Small constant added for numerical stability during weight calculations. A suggested value would be 1e-8.

  • pareto_smoothing: bool

    Whether to apply Pareto smoothing to the importance weights. Normally it is better to set this value as True as it is seen to give better results.