jesterTOV.inference.likelihoods.rex.REXLikelihood

class REXLikelihood(experiment_name, posterior)

Bases: LikelihoodBase

Likelihood function for PREX or CREX neutron skin measurements.

This likelihood constrains the symmetry energy parameters (E_sym, L_sym) using experimental measurements of neutron skin thickness from parity-violating electron scattering. The experimental data is represented as a kernel density estimate (KDE) of the posterior distribution in (E_sym, L_sym) parameter space, which is then evaluated for each candidate EOS.

The neutron skin thickness is primarily sensitive to the pressure of neutron matter at subsaturation densities, which is controlled by E_sym (symmetry energy at saturation) and L_sym (its density slope). These experiments provide complementary information to astrophysical mass-radius measurements, as they probe different density regimes and physical processes.

Parameters:

• experiment_name (str) – Name of the experiment providing the constraint. Must be either “PREX” (Lead Radius Experiment, Pb-208) or “CREX” (Calcium Radius Experiment, Ca-48). The two experiments probe different mass regions and have different systematic uncertainties.

• posterior (Any) – Kernel density estimate of the experimental posterior distribution in (E_sym, L_sym) parameter space. This should be a callable object with a logpdf method that accepts a 2D array of shape (2, n_samples) and returns log-probability values. Typically constructed using scipy.stats.gaussian_kde or similar.

Variables:

• experiment_name (str) – The experiment name (“PREX” or “CREX”)

• counter (int) – Internal counter tracking the number of likelihood evaluations. Used for debugging and performance monitoring.

• posterior (Any) – The KDE object representing the experimental posterior

Raises:

AssertionError – If experiment_name is not “PREX” or “CREX”

Notes

The likelihood evaluation extracts only the E_sym and L_sym parameters from the full parameter dictionary and evaluates the KDE at that point. Other nuclear parameters (K_sat, Q_sat, etc.) do not directly enter this likelihood, though they affect the overall EOS and may have indirect correlations.

The KDE is evaluated in log-space to avoid numerical underflow for low-probability regions and to match the log-likelihood framework used throughout the inference.

See also

ChiEFTLikelihood

Low-density constraints from chiral effective field theory

Examples

Create a PREX likelihood with a pre-computed KDE:

>>> from scipy.stats import gaussian_kde
>>> import numpy as np
>>> # Load PREX posterior samples (example)
>>> samples = np.load("prex_samples.npy")  # shape: (2, n_samples) for (E_sym, L_sym)
>>> kde = gaussian_kde(samples)
>>> from jesterTOV.inference.likelihoods import REXLikelihood
>>> likelihood = REXLikelihood("PREX", kde)
>>> params = {"E_sym": 32.0, "L_sym": 60.0}
>>> log_like = likelihood.evaluate(params, data={})

__init__(experiment_name, posterior)

Methods

__init__(experiment_name, posterior)
evaluate(params)	Evaluate the log-likelihood for PREX/CREX constraints.

Attributes

data	The data for the likelihood.
model	The model for the likelihood.
experiment_name
counter
posterior

counter: int

evaluate(params)

Evaluate the log-likelihood for PREX/CREX constraints.

Parameters:

params (dict[str, Float | Array]) – Dictionary containing EOS parameters. Required keys: - “E_sym” : Symmetry energy at saturation density (MeV) - “L_sym” : Slope of symmetry energy (MeV) Other parameters in the dict are ignored.

Return type:

Float

Returns:

Float – Natural logarithm of the likelihood. This is the log-probability density of the (E_sym, L_sym) point under the experimental posterior KDE.

Notes

The method extracts E_sym and L_sym from the parameter dictionary, constructs a 2D array [E_sym, L_sym], and evaluates the KDE’s logpdf method. The result is a scalar log-probability suitable for combination with other log-likelihoods in the inference pipeline.

experiment_name: str

posterior: Any