Crate bids_modeling 

BIDS Statistical Models implementation.

This crate implements the BIDS-StatsModels specification for defining reproducible neuroimaging analysis pipelines as declarative JSON documents. It corresponds to PyBIDS’ bids.modeling module.

Overview

A BIDS Stats Model defines a directed acyclic graph (DAG) of analysis nodes, each operating at a specific level of the BIDS hierarchy (run, session, subject, dataset). Data flows from lower levels to higher levels through edges, with contrasts propagating upward.

Components

• StatsModelsGraph — The top-level model graph loaded from a JSON file. Validates structure, wires edges between nodes, and executes the full analysis pipeline. Can export to Graphviz DOT format.

• StatsModelsNode — A single analysis node with a statistical model specification, variable transformations, contrasts, and dummy contrasts. Nodes group data by entity values and produce StatsModelsNodeOutputs.

• TransformSpec and apply_transformations() — The pybids-transforms-v1 transformer implementing Rename, Copy, Factor, Scale, Threshold, Select, Delete, Replace, Split, Concatenate, Orthogonalize, Lag, and more.

• HrfModel, spm_hrf(), glover_hrf() — Hemodynamic response function kernels (SPM and Glover canonical forms with optional time and dispersion derivatives). Uses a pure-Rust gammaln implementation matching SciPy’s cephes to machine epsilon.

• auto_model() — Automatically generates a BIDS Stats Model JSON for each task in a dataset, with Factor(trial_type) at the run level and pass-through nodes at higher levels.

• GlmSpec, MetaAnalysisSpec — Statistical model specifications with design matrix construction, VIF computation, and formatted output.

Example

use bids_modeling::StatsModelsGraph;

let mut graph = StatsModelsGraph::from_file(
    std::path::Path::new("model-default_smdl.json")
).unwrap();
graph.validate().unwrap();
println!("DOT graph:\n{}", graph.write_graph());

// Execute the model
let outputs = graph.run();
for output in &outputs {
    println!("Node: {}, Contrasts: {}", output.node_name, output.contrasts.len());
}

Re-exports

pub use auto_model::auto_model;

pub use graph::StatsModelsGraph;

pub use hrf::HrfModel;

pub use hrf::compute_regressor;

pub use hrf::glover_hrf;

pub use hrf::spm_hrf;

pub use node::ContrastInfo;

pub use node::StatsModelsEdge;

pub use node::StatsModelsNode;

pub use node::StatsModelsNodeOutput;

pub use node::build_groups;

pub use spec::GlmSpec;

pub use spec::MetaAnalysisSpec;

pub use spec::Term;

pub use spec::compute_vif;

pub use spec::dummies_to_vec;

pub use spec::format_correlation_matrix;

pub use spec::format_design_matrix;

pub use transformations::TransformSpec;

pub use transformations::TransformerManager;

pub use transformations::apply_transformations;

pub use transformations::expand_wildcards;

Modules

auto_model

Automatic BIDS Stats Model generation.

graph

BIDS-StatsModels directed acyclic graph.

hrf

Hemodynamic response functions for fMRI statistical modeling.

node

Analysis nodes and edges in a BIDS-StatsModels graph.

spec

Statistical model specifications for BIDS-StatsModels.

transformations

Variable transformations from the pybids-transforms-v1 specification.