Package index • dkge

Workflow — start here

Primary functions for fitting, contrasting, classifying, and predicting

dkge(): Fit DKGE across multiple subjects
dkge_contrast(): Compute DKGE contrasts with cross-fitting
dkge_classify(): Cross-validated classification on DKGE effect patterns
dkge_infer(): Unified inference for DKGE contrasts
dkge_predict(): Predict DKGE contrasts for new subjects (out-of-sample)
dkge_freeze(): Freeze a DKGE fit into a compact model for prediction
dkge_bootstrap_analytic(): Analytic first-order bootstrap in the design space
dkge_bootstrap_projected(): Subject-level projection bootstrap in medoid space
dkge_bootstrap_qspace(): Multiplier bootstrap in the design space (q-space)

Advanced — model construction

Lower-level functions for custom workflows

dkge_fit(): Fit a Design-Kernel Group Embedding (DKGE) model
dkge_fit_from_input(): Fit DKGE from an input descriptor
dkge_fit_from_kernels(): Fit DKGE from precomputed subject effect kernels
dkge_weights(): Create a DKGE voxel-weight specification
dkge_weights_auto(): Default DKGE weight specification
dkge_weights_prior_mask(): Build prior weights from a mask
dkge_weights_prior_roi(): Build prior weights from ROI labels
design_kernel() dkge_design_kernel(): Build a flexible design-similarity kernel
dkge_mapper(): Create a pluggable DKGE anchor mapper for dense rendering
dkge_mapper_spec(): Specify a DKGE mapper strategy for the transport pipeline
dkge_transport_spec(): Transport specification helper
dkge_transport_contrasts_to_medoid(): Transport subject contrasts to a medoid parcellation
dkge_transport_loadings_to_medoid(): Transport component loadings to a medoid parcellation
dkge_transport_to_medoid_sinkhorn() dkge_transport_to_medoid_sinkhorn_cpp(): Transport cluster values to a medoid via entropic Sinkhorn OT
dkge_transport_to_voxels(): Transport DKGE quantities directly to voxel space
dkge_transport_service(): Construct a transport service
dkge_classification_spec(): Classification specification helper
dkge_inference_spec(): Inference specification helper
dkge_define_folds(): Define folds for K-fold cross-validation
dkge_subject(): Construct a DKGE subject record
dkge_data(): Bundle subject-level inputs for DKGE
dkge_prepare_transport(): Prepare subject-to-medoid transport operators for reuse
dkge_pipeline(): End-to-end DKGE workflow
dkge_loso_contrast(): Leave-one-subject-out DKGE contrast
dkge_analytic_loso(): Analytic LOSO contrast using eigenvalue perturbation
dkge_contrast_validated(): Dual-path DKGE contrasts with coverage diagnostics
dkge_cv_kernel_grid(): LOSO kernel grid search
dkge_cv_kernel_rank(): Combined kernel and rank selection via pre-screening and LOSO CV
dkge_cv_rank_loso(): LOSO cross-validation for rank selection
dkge_cv_train_latent_classifier(): Cross-fitted linear classifiers in the DKGE latent space
dkge_pooled_cov_q(): Pooled design-space covariance and Cholesky factor
dkge_predict_loadings(): Predict DKGE loadings for new subjects (out-of-sample)
dkge_predict_stream(): Streaming prediction for new subjects via a loader
dkge_predict_subjects(): Convenience prediction for subject collections
dkge_preprocess_blocks(): Preprocess multiple blocks into the DKGE training space
dkge_project_btil() dkge_project_block(): Project DKGE data into component space
dkge_project_blocks(): Project new blocks into DKGE score space
dkge_project_cluster(): Project a new cluster/voxel vector onto DKGE components
dkge_project_clusters(): Project multiple cluster/voxel vectors
dkge_project_clusters_to_latent(): Project subject clusters into the DKGE latent space
dkge_sim_toy(): Simulate toy DKGE datasets with known factorial structure
dkge_update_weights(): Refit a DKGE object with a new voxel-weight specification

Anchors and rendering

Functions for building and applying spatial anchor mappers

dkge_build_renderer(): Prepare reusable rendering objects for a fitted DKGE model
dkge_render_subject_values(): Render per-subject values to anchors and voxels
dkge_build_anchor_kernels(): Fold-aware anchor kernel construction
dkge_make_anchors(): Build or validate anchor coordinates in MNI space
dkge_anchor_aggregate(): Aggregate anchor fields with optional Laplacian smoothing
dkge_anchor_contrast_from_direction(): Build an anchor contrast from a feature-space direction
dkge_anchor_contrast_from_prototypes(): Build an anchor contrast from prototype feature sets
dkge_anchor_diagnostics(): Extract anchor diagnostics from a DKGE fit
dkge_anchor_fit(): Fit DKGE using feature-anchored subject kernels
dkge_anchor_graph(): Construct a kNN anchor graph and Laplacian
dkge_anchor_targets_from_directions(): Assemble anchor targets from feature-space directions
dkge_anchor_targets_from_prototypes(): Assemble anchor targets from prototype sets
dkge_anchor_to_voxel_apply(): Decode anchor values to voxel space
dkge_anchor_to_voxel_fit(): Fit a sparse anchor-to-voxel decoder
dkge_input_anchor(): Anchor-based DKGE input descriptor
fit_mapper(): Fit a mapper on subject/reference features
predict_mapper(): Apply a fitted mapper to new source values
apply_mapper(): Apply a fitted mapper to values

Extension points — S3 generics

Generics for integrating custom kernels, folds, and inputs

as_dkge_kernel(): Convert to a DKGE design kernel
as_dkge_folds(): Convert to DKGE fold assignments
dkge_run_service(): Run a dkge service object
dkge_contrast_service(): Construct a contrast service
dkge_inference_service(): Construct an inference service

Kernel and alignment utilities

Design kernels, K-Procrustes alignment, and kernel diagnostics

kernel_alignment() dkge_kernel_alignment(): Kernel alignment score
kernel_roots() dkge_kernel_roots(): Robust kernel roots
dkge_kernel_prescreen(): Kernel alignment pre-screening
dkge_k_orthonormalize(): Robust K-orthonormalization
dkge_procrustes_K(): K-orthogonal Procrustes alignment
dkge_align_bases_K(): Align multiple bases to a reference
dkge_consensus_basis_K(): Consensus K-orthonormal basis (K-Procrustes mean)
dkge_projector_K(): K-orthogonal projector onto span(T) in effect space
dkge_cosines_K(): Principal-angle cosines in the K-metric
dkge_jd_control(): Control parameters for DKGE joint diagonalisation
dkge_align_effects(): Align effect kernels across subjects with partial overlap
dkge_target_factor(): Target helper for a single factor
dkge_target_interaction(): Target helper for an interaction
dkge_targets(): Build classification targets from a DKGE fit

Cluster and CPCA utilities

Cluster-level projections and CPCA filtering

dkge_cpca_fit(): Fit DKGE with CPCA filtering
dkge_fit_cpca(): Fit DKGE bases on CPCA-filtered components
dkge_cpca_split_chat(): Split compressed covariance into design/residual parts (CPCA inside-span)
dkge_cluster_betas(): Compute cluster-level betas from neuroim2 objects
dkge_cluster_loadings(): Cluster-to-latent loadings for DKGE subjects
dkge_cluster_ts(): Aggregate voxel time series into cluster means
dkge_regress(): Multi-output regression on DKGE effects
dkge_transform_block(): Preprocess a subject block into DKGE training space
dkge_neuro_loader(): Build a streaming loader backed by neuroim2 objects

Visualization and diagnostics

Plotting and metric functions

dkge_plot_effect_loadings(): Effect-space loadings heatmap (K %*% U)
dkge_plot_info_anchor(): Anchor-level information plots
dkge_plot_scree(): DKGE scree plot with cumulative curve
dkge_plot_subject_contrib(): Subject weights and per-component energy heatmap
dkge_plot_subspace_stability(): Subspace stability via principal angles
dkge_plot_suite(): DKGE "Five Fundamentals" dashboard
dkge_diagnostics(): Summarise DKGE diagnostics
dkge_component_stats() dkge_write_component_stats(): Component-level consensus statistics
dkge_variance_explained(): Compute per-component variance explained
dkge_info_map_from_classifier(): Decoder-style information map from latent classifier weights
dkge_info_map_haufe(): Haufe-style encoding maps from latent classifiers
dkge_info_map_loco(): Group-LOCO anchor importance (zeroing proxy)
dkge_confusion(): Fold-wise confusion matrices for DKGE classification
dkge_paint_medoid_map(): Paint medoid cluster values back to a label volume
theme_dkge(): DKGE minimal theme for ggplot2 outputs

Internal building blocks

Low-level functions exposed for advanced use and testing

dkge_freedman_lane(): Freedman-Lane permutations for DKGE (scaffold)
dkge_signflip_maxT(): One-sample sign-flip max-T inference on transported subject maps
dkge_one_se(): One standard-error rule selection helper
dkge_component_stats() dkge_write_component_stats(): Component-level consensus statistics
dkge_write_group_map(): Write a group map as NIfTI using a medoid label image
dkge_clear_sinkhorn_cache(): Clear cached dual variables for Sinkhorn warm-starts
helmert_contrasts(): Helmert contrasts for a set of factors
sum_contrasts(): Sum-to-zero contrasts for a set of factors
dkpp_select_anchors(): Determinantal k-means++ (d-kpp) anchor selection
pairwise_sqdist_cpp(): Pairwise squared distances between two point sets (rows)
predict(<dkge>) predict(<dkge_model>): Predict contrasts for new subjects using a DKGE fit
print(<dkge_contrasts>): Print method for dkge_contrasts
print(<dkge_folds>): Print method for dkge_folds
print(<dkge_inference>): Print method for dkge_inference
as.data.frame(<dkge_inference>): Convert DKGE inference results to a tidy data frame
as.matrix(<dkge_contrasts>): Extract contrast values as matrix
dkge-info-maps: DKGE classifier information maps
dkge-latent-utils: DKGE latent-space utilities