Model Parameters

Class providing model parameters for Bayesian inference—the individual free parameters of a Galacticus model that are explored during parameter estimation (e.g.via MCMC). Each parameter has a name, a prior distribution (with logPrior, priorSample, priorInvert, priorMinimum, and priorMaximum methods), and a bijective mapping to an unconstrained real line for efficient sampling (via map/unmap). Implementations include active parameters that vary during inference, and fixed parameters held at constant values.

Default implementation: modelParameterActive

Methods

name → type(varying_string)

Return the name of this parameter as it appears in the Galacticus parameter file and in output metadata, used to identify the parameter when applying posterior sampler updates.

logPrior → double precision

Return the natural logarithm of the prior probability density evaluated at the physical parameter value x, used in computing the log-posterior during Bayesian inference.

• double precision, intent(in ) :: x

priorSample → double precision

Draw a random sample from the prior distribution for this parameter, returning a physical parameter value; used to initialize the posterior sampler or generate prior predictive samples.

priorInvert → double precision

Invert the prior, returning the parameter value given the cumulative probability.

• double precision, intent(in ) :: f

priorMinimum → double precision

Return the minimum non-zero value of the prior for this parameter.

priorMaximum → double precision

Return the maximum non-zero value of the prior for this parameter.

randomPerturbation → double precision

Return a random perturbation for this parameter.

map → double precision

Apply the bijective mapping to transform the physical parameter value x onto the unconstrained real line used by the posterior sampler (e.g.logarithm for positive-definite parameters).

• double precision, intent(in ) :: x

unmap → double precision

Apply the inverse bijective mapping to transform the sampler’s unconstrained variable x back to the physical parameter value used to evaluate the model.

• double precision, intent(in ) :: x

mapJacobian → double precision

Compute the Jacobian of the map at the given parameter value.

• double precision, intent(in ) :: x

modelParameterActive

An active model parameter class representing a free parameter that participates in posterior sampling, with a prior distribution, perturbation distribution, and optional mapping transformation. The parameter name is set by [name], with prior and perturbation distributions provided as objects, and a unary operator applied to transform the parameter value for model evaluation.

(Default implementation)

Methods

• isSlow — Return true if changes in this paper may lead to slow likelihood evaluation.

Parameters

• [name] (default var_str('')) — The name of the active model parameter as it appears in the Galacticus parameter file and output metadata; used to identify and retrieve the parameter during posterior sampling runs.

• [slow] (default .false.) — If true, changes in the parameter are considered to result in slow likelihood evaluations.

modelParameterDerived

A model parameter class in which the parameter value is not sampled directly but instead derived algorithmically from other model parameters via a mathematical expression. The parameter name is set by [name] and the derivation formula by [definition], enabling computed quantities to be tracked as named parameters during posterior sampling.

Methods

• definition — Return the definition for this parameter.

• isInteger — Return true if the parameter is an integer.

Parameters

• [name] — The name of the derived model parameter as it appears in the Galacticus parameter file and posterior sampling output; used to identify this computed quantity in the parameter chain.

• [definition] — The mathematical expression or formula used to derive this parameter’s value from other model parameters during posterior sampling.

• [isInteger] (default .false.) — If true the derived value is treated as an integer, otherwise it is considered to be a float.

modelParameterInactive

An inactive model parameter class representing a fixed parameter that does not participate in posterior sampling and has no prior distribution or perturbation applied during inference. The parameter name is set by [name], making it available as a named parameter in the model parameter set without affecting the sampling space.

Parameters

• [name] (default var_str('')) — The name of the inactive model parameter as it appears in the Galacticus parameter file; held at a fixed value during posterior sampling without contributing to the sampler’s free parameter space.