rxmc.evidence.Evidence#

class rxmc.evidence.Evidence(constraints: list[Constraint] | None = None, parametric_constraints: list[Constraint] | None = None, weights: ndarray = None, weights_parametric: ndarray = None)[source]#

Bases: object

A collection of independent constraints sharing a common physical model.

Each Constraint represents a set of observations paired with a likelihood model. All constraints must share the same physical-model parameters, but may have different likelihood-model parameters.

Optional per-constraint weights scale the contribution of each constraint to the total log likelihood.

Parameters:

  • constraints (list of Constraint, optional) – Constraints whose likelihood models have no free parameters.

  • parametric_constraints (list of Constraint, optional) – Constraints whose likelihood models have free parameters.

  • weights (np.ndarray, optional) – 1-D array of weights for constraints. Defaults to all ones.

  • weights_parametric (np.ndarray, optional) – 1-D array of weights for parametric_constraints. Defaults to all ones.

Raises:

  • ValueError – If both constraints and parametric_constraints are empty.

  • ValueError – If any constraint uses different physical-model parameters than the first.

  • ValueError – If a non-parametric constraint appears in parametric_constraints, or vice versa.

  • ValueError – If the number of data points is less than the number of free parameters (under-constrained model).

  • ValueError – If weights or weights_parametric do not match the corresponding list length.

__init__(constraints: list[Constraint] | None = None, parametric_constraints: list[Constraint] | None = None, weights: ndarray = None, weights_parametric: ndarray = None)[source]#

Methods

__init__([constraints, ...])

log_likelihood(model_params[, likelihood_params])

Weighted sum of log likelihoods over all constraints.
log_likelihood(model_params, likelihood_params: list[tuple] | None = None)[source]#

Weighted sum of log likelihoods over all constraints.

Parameters:

  • model_params (tuple) – Parameters of the physical model.

  • likelihood_params (list of tuple, optional) – One tuple of likelihood parameters per entry in parametric_constraints. Defaults to an empty list.

Returns:

float – Total weighted log likelihood.