Package index • kernR

Kernel primitives

Kernel specifications, Gram matrices, bandwidth selection, and the weighted HSIC statistic.

kernel_spec(): Create a Kernel Specification
kernel_matrix(): Compute a Kernel Matrix
select_bandwidth(): Select Kernel Bandwidth
resolve_bandwidth(): Resolve Kernel Bandwidth
weighted_hsic_stat(): Weighted HSIC Statistic

Independence and two-sample tests

HSIC independence test and MMD two-sample test, with low-rank acceleration.

hsic_test(): HSIC Independence Test
mmd_test(): MMD Two-Sample Test
hsic_test_nystrom(): HSIC Independence Test via Low-Rank Factorisation
plot(<kernel_test_result>): Plot a Kernel Test Result

Goodness-of-fit, concordance, and calibration

One-sample kernel Stein discrepancy, a k-sample concordance test over ensembles, and a graded coverage / calibration diagnostic, with low-rank acceleration.

ksd_test(): Kernel Stein Discrepancy Goodness-of-Fit Test
ksd_test_nystrom(): Accelerated Kernel Stein Discrepancy Goodness-of-Fit Test (Nystrom)
concordance_test(): Kernel k-sample Concordance Test
concordance_test_nystrom(): Accelerated Kernel k-sample Concordance Test (Nystrom / RFF)
coverage_test(): Coverage / Calibration Diagnostic for a Predictive Ensemble
joint_coverage_test(): Joint (Multivariate) Calibration Diagnostic for a Predictive Ensemble
calibration_suite(): Calibration suite for a continuous prediction
gaussian_score(): Score function for a multivariate normal target
numeric_score(): Finite-difference score from a log-density

Causal association (bd-HSIC)

Backdoor-adjusted HSIC for testing causal association after confounder adjustment.

bd_hsic_test(): Backdoor-HSIC Test for Causal Association

Mechanism-consistency test

Mechanism-consistency testing against a calibrated process model, with grounding labelled per the Independent Oracle Principle.

mechanism_consistency_test(): Mechanism-consistency test

Assimilative causal inference (ACI)

The relative-entropy causal metric and the threshold-free causal influence range; consumed by kalmix.

relative_entropy(): Relative Entropy Between Two Gaussians (the ACI causal metric)
relative_entropy_ensemble(): Relative Entropy Between Two Sample Ensembles (Gaussian approximation)
cir_objective(): Objective Causal Influence Range (threshold-free)

Distributional treatment effects

Doubly robust kernel tests for distributional treatment effects, including a PESTO-scenario convenience.

dr_date_test(): Doubly Robust Distributional Average Treatment Effect Test (DR-DATE)
dr_dett_test(): Doubly Robust Distributional Effect on the Treated Test (DR-DETT)
dr_date_scenario(): DR-DATE for Two PESTO Ensemble Scenarios
kernel_causal_test(): Unified Kernel Causal Test

Hierarchical and clustered designs

Within-cluster permutation for nested data (farms, hospitals, schools).

hierarchical_test(): Hierarchical Kernel Causal Test

Sensitivity and identifiability

First- and total-order kernel sensitivity; pre-IES identifiability screening.

hsic_identifiability(): HSIC-Based Identifiability Diagnostic
hsic_sensitivity(): HSIC-Based Distributional Sensitivity Index
plot(<hsic_identifiability>): Plot an HSIC Identifiability Scan
plot(<hsic_sensitivity>): Plot HSIC-Sensitivity Indices

Density-ratio and propensity

Plug-in density-ratio estimation with four backends; propensity-score diagnostics.

fit_density_ratio(): Fit a Density-Ratio Model
predict_density_ratio(): Predict from a Fitted Density-Ratio Model
estimate_density_ratio(): Estimate Density Ratios (backwards-compatible wrapper)
estimate_propensity(): Estimate Propensity Scores
assess_overlap(): Assess Propensity Score Overlap
plot_weights(): Plot Weight Diagnostics
effective_sample_size(): Compute Effective Sample Size

Low-rank acceleration

Nyström and Random Fourier Feature factorisations for large-n kernels.

nystrom_factor(): Nystrom Low-Rank Kernel Factorisation
rff_features(): Random Fourier Features for the RBF Kernel

Kernel downscaling and distribution regression

Conditional mean embedding, distribution regression, and aggregate downscaling.

kernel_downscale(): Kernel-Based Statistical Downscaling
fit_cme(): Estimate Conditional Mean Embedding via Kernel Ridge Regression
dist_regression(): Kernel Distribution Regression
aggregate_downscale(): Aggregate-Likelihood Downscaling
posterior_sample_aggregate(): Sample from the posterior of an aggregate-downscale fit
predict(<cme_fit>): Predict Conditional Mean Embedding Weights at New Points
print(<cme_fit>): Print a Conditional Mean Embedding Fit
predict(<dist_regression>): Predict from a Fitted Distribution Regression Model

Posterior-predictive check

MMD-based posterior-predictive check; consumes the PESTO cross-package contract.

mmd_ppc(): MMD Posterior-Predictive Check
pesto_ensemble(): PESTO Ensemble Manifest (Constructor)

Orchestra manifest contract

Emitting a kernR verdict as the federation’s shared, integrity-hashed manifest, and verifying one.

as_orchestra_manifest(): Emit a kernR verdict as an orchestra manifest
orchestra_manifest: The orchestra ensemble-manifest contract (kernR-side implementation)
verify_manifest(): Verify an orchestra manifest's payload integrity

Tidiers

broom-style one-row-per-term summaries of kernR verdict objects.

tidy(<kernel_test_result>): Tidy a kernR Test Result Into a One-Row-Per-Term Data Frame
tidy(<mechanism_consistency_result>): Tidy a Mechanism-Consistency Result

Design

Latin-hypercube design helper.

lhs_design(): Latin-Hypercube Design Over Bounded Parameters

Typed declines

The classed abstention kernR returns instead of a verdict when the effective sample size is below the floor.

is_kernR_abstention(): Did a kernR result decline to stand behind its own verdict?