pymargins.WrappedFDAdapter

class pymargins.WrappedFDAdapter

Base for models with smooth predict but no exposed analytical derivative.

Uses _gradients.make_predict_with_fd_jvp to wrap the framework’s native predict with a JAX-compatible custom JVP. FD is hidden inside the JVP; downstream autodiff over the estimand structure remains exact.

Concrete subclasses provide native_predict(beta_np, X) and the wrapper is constructed automatically.

__init__()

Methods

__init__()
attach(session)	Attach this adapter to a Margins session.
bootstrap_state()	Replay state for a refitted adapter.
coefficients()	Return β̂ as a 1D JAX array.
column_index_of_variable(name)	Return the design-matrix column index corresponding to a variable.
covariance([vcov_spec])	Return Σ̂ as a 2D JAX array.
design_matrix_from_df(df)	Build a design matrix from a concrete DataFrame of evaluation rows.
native_predict(beta_np, X)	Framework-native predict.
predict(beta, X[, offset])	FD-wrapped predict.
refit(resampled_data, *[, index])	Refit the model on resampled data, returning a new adapter.
variable_metadata()	Return per-variable metadata used by averaging and validation.

Attributes

gradient_backend_recommendation	Recommend wrapped_fd because predict uses finite-difference JVP.
n_outcomes	Number of outcome classes for multi-outcome models, default 1.
outcome_labels	Outcome class labels for multi-outcome models, or None.
supported_inference_methods	Wrapped-FD adapters support delta, simulation, and bootstrap.
supports_jax_autodiff	FD-based JVP is not native autodiff; flag for diagnostics.
training_data	The training data used to fit the model.