Package index • kagu

Model

Main user-facing class.

KaguModel: KaguModel - Bayesian graphical causal model

Mechanisms

Probabilistic models for node conditional distributions.

Mechanism: Mechanism abstract base class
GPMechanism: Gaussian-process mechanism (default)

Effects

Causal effect estimation and results.

EffectResult: Result of a causal effect query
compute_effect(): Compute the causal effect of source on target

Structure discovery

Posterior over DAGs via marginal-likelihood model comparison.

kagu_discover(): Discover causal structure from data
DiscoveryResult: Result of a causal structure search
enumerate_dags(): Enumerate all DAGs over a set of nodes
kagu_plot_discovery(): Plot the posterior distribution over DAGs

DAG utilities

Graph operations on DAG specifications.

validate_dag(): Validate a DAG specification
topological_sort(): Topological sort (Kahn's algorithm)
ancestors(): Ancestors of a node
descendants(): Descendants of a node
is_ancestor(): Test whether one node is an ancestor of another
node_depth(): Depth of each node in the DAG

Plots

kagu_plot_dag(): Plot the DAG structure
kagu_plot_posterior(): Plot a fitted node's summary terms

Save / load

kagu_save(): Save a fitted model to disk
kagu_load(): Load a model saved with kagu_save()

Fitting

fit_node(): Fit a single node's conditional distribution

Summaries

build_summary_table(): Build a per-node summary table across all nodes

Internal

File-level documentation pages.

dag: DAG validation and graph utilities
discover: Causal structure discovery
effects: Causal effect estimation via the do-operator
inference: Per-node model fitting
io: Save and load fitted KaguModel objects
mechanisms: Mechanism base class and the Gaussian-process implementation
plots: DAG and posterior visualisation
summary: Summary table for a fitted KaguModel