causalcompass.algorithms.PCMCI

class causalcompass.algorithms.PCMCI(tau_max=3, pc_alpha=0.05, alpha=0.05, cond_ind_test=None, seed=None)[source]

Constraint-based causal discovery method that combines the PC algorithm with Momentary Conditional Independence tests for time series data.

References

https://github.com/jakobrunge/tigramite

Parameters:

  • tau_max (int, default 3) – Maximum time lag

  • pc_alpha (float, default 0.05) – Significance level for PC stable step

  • alpha (float, default 0.05) – Significance level for MCI test

  • cond_ind_test (CondIndTest or None, default None) – Conditional independence test object from tigramite.independence_tests or an external test passed as a callable based on tigramite.independence_tests.CondIndTest. If None, ParCorr() will be used by default in run()

Examples

>>> from causalcompass.algorithms import PCMCI
>>> model = PCMCI(tau_max=3, pc_alpha=0.05, alpha=0.05)
>>> predicted_adj = model.run(X)
>>> all_metrics, no_diag_metrics = model.eval(true_adj, predicted_adj)
__init__(tau_max=3, pc_alpha=0.05, alpha=0.05, cond_ind_test=None, seed=None)[source]

Initialize PCMCI

Methods

__init__([tau_max, pc_alpha, alpha, ...])

Initialize PCMCI

eval(true_adj, predicted_adj[, shd_thresholds])

Evaluate the predicted adjacency matrix against the ground truth.

run(X)

Run PCMCI algorithm.

run_raw(X)

Run PCMCI once and return the raw p-value matrix.

run_threshold_sweep(X, thresholds)

Run the algorithm once and post-process the raw result for each threshold.