Module predict

Structs§

BernoulliMarginalSlopePredictor
BinomialLocationScalePredictor: Binomial location-scale predictor: two blocks (threshold + log-sigma).
CoefficientUncertaintyResult
GaussianLocationScalePredictor: Gaussian location-scale predictor: two blocks (mean + log-sigma).
PredictInput: Input to the prediction trait. Contains the design matrix and metadata needed for point prediction + uncertainty quantification.
PredictPosteriorMeanResult
PredictResult
PredictUncertaintyOptions
PredictUncertaintyResult
PredictionWithSE: Point prediction with optional standard errors on the linear predictor.
StandardPredictor: Standard (single-block) GAM predictor.
SurvivalPredictor
TrainingSupport: Per-axis training support range used by boundary and OOD corrections. For each predictor axis we record the empirical [min, max] from training. Boundary correction inflates variance for x_i within a small fraction of the range from either edge; OOD inflation inflates variance for x_i outside [min, max] proportional to (excess / range).
TransformationNormalPredictor: Predictor for transformation-normal (PIT) models.

PredictableModel: Trait for models that can produce predictions from new data.

coefficient_uncertainty: Coefficient-level uncertainty and confidence intervals.
coefficient_uncertaintywith_mode: Coefficient-level uncertainty and confidence intervals with explicit covariance mode.
enrich_posterior_mean_bounds: Compute and attach TransformEta confidence bounds to a posterior-mean result.
predict_gam: Generic engine prediction for external designs. This API is domain-agnostic: callers provide only design matrix, coefficients, offset, and family.
predict_gam_posterior_mean: Nonlinear posterior-mean prediction with coefficient uncertainty propagation.
predict_gam_posterior_meanwith_backend
predict_gam_posterior_meanwith_fit: Nonlinear posterior-mean prediction with link-state support for SAS/mixture families.
predict_gamwith_uncertainty: Prediction with coefficient uncertainty propagation.
se_from_covariance: Compute standard errors from a covariance matrix (sqrt of diagonal).