Module generative 

Structs

GenerativeSpec

First-class generative specification: mean process + observation noise.

Enums

NoiseModel

Observation-noise model used for generative sampling.

Traits

CustomFamilyGenerative

Extension trait for custom multi-block families that provide explicit generative semantics (mean + observation noise) at a fitted state.

Functions

family_noise_parameter

THE single source of truth for the scalar dispersion the generative observation model uses for a fitted family — the value handed to NoiseModel::from_likelihood / generativespec_from_predict as gaussian_scale.

generativespec_from_predict

Build a generative specification for built-in GAM families from eta/mean.

sampleobservation_replicate_chunks

Draw replicate chunks in deterministic draw order without materializing the full n_draws × nobs matrix.

sampleobservation_replicates

Collect multiple synthetic replicates into an n_draws × nobs matrix.

sampleobservation_seeded_replicate_chunks

Draw a seekable range of independently seeded replicate chunks.

sampleobservation_seeded_replicates

Collect a seekable range into an allocating n_draws × nobs matrix.

sampleobservations

Draw one synthetic observation vector from a generative spec.