bedroc.isotope_anomalies package

Submodules

bedroc.isotope_anomalies.core module

Helper functions and classes for nucleosynthetic isotope anomalies

bedroc.isotope_anomalies.core.DATA: Traversable = PosixPath('/home/docs/checkouts/readthedocs.org/user_builds/bedroc/checkouts/latest/bedroc/isotope_anomalies/NC_CC_AllData.csv')

Isotope anomalies data file

class bedroc.isotope_anomalies.core.GroupData(name: str, datapath: Traversable | Path = PosixPath('/home/docs/checkouts/readthedocs.org/user_builds/bedroc/checkouts/latest/bedroc/isotope_anomalies/NC_CC_AllData.csv'), chondrites: Iterable[str] | None = None, elements: Iterable[str] | None = None, label_offsets: dict[str, tuple[float, float]] | None = None, eigvec_label_offsets: dict[str, tuple[float, float]] | None = None)

Bases: object

Grouped isotope data

Parameters:

  • name – Name of the group

  • datapath – Path to the data file. Defaults to DATA.

  • chondrites – Chondrites to select. Defaults to None to select all.

  • elements – Elements to select. Defaults to None to select all.

  • label_offsets – Offsets for plotting the feature (isotope) labels. Defaults to None.

  • eigvec_label_offsets – Offsets for plotting the eigenvectors. Defaults to None.

name: str
datapath: Traversable | Path = PosixPath('/home/docs/checkouts/readthedocs.org/user_builds/bedroc/checkouts/latest/bedroc/isotope_anomalies/NC_CC_AllData.csv')
chondrites: Iterable[str] | None = None
elements: Iterable[str] | None = None
label_offsets: dict[str, tuple[float, float]] | None = None
eigvec_label_offsets: dict[str, tuple[float, float]] | None = None
data: DataContainer
property idata: DataTree

Inference data

property model: Model

PyMC model

to_excel(directory: Path = PosixPath('.')) → Path

Exports the inference data to Excel.

Parameters:

directory – Directory to export to. Defaults to the current directory.

Returns:

Excel file path

to_pickle(directory: Path = PosixPath('.')) → Path

Exports the inference data to a pickle file.

Parameters:

directory – Directory to export to. Defaults to the current directory.

Returns:

Pickle file path

run_bayesian_pca(random_seed: int | None = None) → None

Runs the Bayesian PCA

Parameters:

random_seed – Optional random seed

plot_pca(ax: Axes, plot_legend: bool = True, include_title: bool = True, skip: int = 10, plot_eigenvectors: bool = True) → Axes

Plots the Bayesian and deterministic PCA.

Parameters:

  • ax – Axis

  • plot_legend – Plots the legend. Defaults to True.

  • include_title – Adds a title. Defaults to True.

  • skip – Take every skip-th sample. Defaults to 10.

  • plot_eigenvectors – Plot and label the eigenvectors. Defaults to True.

Returns:

Axes

plot_predicted_observations(latent_factor_means: ndarray[tuple[Any, ...], dtype[float64]], latent_factor_stds: ndarray[tuple[Any, ...], dtype[float64]], data_names: list[str], reconstruction_only: bool = False, random_seed: int | None = None) → tuple[list[Figure], DataFrame]

Plots predicted observations from latent factors, with optional noise.

This function supports two related but distinct visualizations:

  1. Noise-free reconstruction (reconstruction_only=True)

    • Uses the latent variables and loadings alpha to compute the mean structure of the predicted data: mu = Z @ alpha.

    • Corresponds to classical PCA-style backprojection, but propagates uncertainty from the latent factors.

    • Use this mode to assess how well the inferred latent structure explains the underlying signal in the observations.

  2. Posterior predictive simulation (reconstruction_only=False)

    • Simulates noisy observations from the generative model’s likelihood (Student-T) for new data points.

    • Accounts for uncertainty in both the latent factors and the observation noise.

    • Since true per-data noise is unknown for new points, the feature-level noise is estimated from the training data as the mean per feature.

    • Draws Student-T random samples consistent with the model’s posterior for the degrees of freedom (nu).

Parameters:

  • latent_factor_means – Means of the latent factors (n_data, n_components)

  • latent_factor_stds – Standard deviations of the latent factors (n_data, n_components)

  • data_names – Data names

  • reconstruction_only –

    • If True, plot only the latent-space reconstruction

    • If False (default), simulate and plot noisy posterior-predictive observations

  • random_seed – Seed for random number generation to enable reproducibility. Defaults to None.

Returns:

  • Figures visualizing the predicted observations

  • Summary statistics (mean, std, quantiles, etc.) of the predicted observations across posterior samples

Return type:

tuple

plot_reconstructed_observations(reconstruction_only: bool = False, random_seed: int | None = None) → tuple[list[Figure], DataFrame]

Plots reconstructions of the observed isotope anomalies.

This function supports two related but distinct visualizations:

  1. Noise-free reconstruction (reconstruction_only=True)

    • Uses the posterior samples of the latent variables Z and loadings alpha to compute the mean structure of the data: mu = Z @ alpha.

    • Corresponds to classical PCA-style backprojection, but with full Bayesian uncertainty.

    • Use this mode to assess how well the inferred latent structure explains the underlying signal in the observations.

  2. Posterior predictive simulation (reconstruction_only=False)

    • Draws noisy observations from the models’ likelihood (Student-t).

    • This evaluates how well the full generative model predicts the measured data, including observational noise.

    • Is the most direct and appropriate comparison to the actual observations (posterior predictive check).

Parameters:

  • reconstruction_only –

    • If True, plot only the latent-space reconstruction

    • If False (default), simulate and plot noisy posterior-predictive observations

  • random_seed – Random seed for posterior predictive sampling. Defaults to None.

Returns:

  • Figures visualizing the reconstructed observations

  • Summary statistics (mean, std, quantiles, etc.) of the reconstructed observations across posterior samples

Return type:

tuple

_get_summary_dataframe(samples: ndarray[tuple[Any, ...], dtype[float64]], data_names: list[str]) → DataFrame

Gets a dataframe of summary statistics for data

Parameters:

  • samples – Data samples (n_data, n_features, n_samples)

  • data_names – Data names (n_data,)

Returns:

Dataframe of summary statistics

_plot_reconstruction(title_prefix: str, samples: ndarray[tuple[Any, ...], dtype[float64]], latent_factor_means: ndarray[tuple[Any, ...], dtype[float64]], data_names: list[str], observed: ndarray[tuple[Any, ...], dtype[float64]] | None = None) → list[Figure]

Helper to plot the reconstructed or predicted observations

Parameters:

  • title_prefix – Prefix of the title

  • samples – Data samples (n_data, n_features, n_samples)

  • latent_factor_means – Means of the latent factors

  • data_names – Data names

  • observed – Observed data to also plot, if not None. Defaults to None.

Returns:

Figures visualizing the reconstruction

bedroc.isotope_anomalies.core.get_group_data() → dict[str, GroupData]

Gets the group data

bedroc.isotope_anomalies.core.is_vesta_group(chondrite_name: str) → bool

Checks if a chondrite is part of the Vesta Group.

Parameters:

chondrite_name – Chondrite name

Returns:

True if part of the Vesta Group, otherwise False

bedroc.isotope_anomalies.core.get_color(chondrite_name: str, reservoir_name: str) → tuple[str, str]

Gets the color and the light color associated with the chondrite and/or reservoir.

Parameters:

  • chondrite_name – Chondrite name

  • reservoir_name – Reservoir name

Returns:

  • Color

  • Light color

Return type:

tuple

Module contents

Nucleosynthetic isotope anomalies

Citation:

Sossi, Paolo A., and Bower, Dan J. (2026), Homogeneous accretion of the Earth in the inner Solar System, Nature Astronomy.