Causal discovery (PCMCI-style)

Primary reference: Runge et al. (2019) Nature Communications 10, 2553.

time_lagged_causality_network() evaluates transfer entropy or Granger causality at multiple lags — a PCMCI-style building block for lagged discovery.

Assumptions

• Panel of aligned univariate series.

• Lagged driver X→Y should dominate reverse flow at the true lag (fixture: pcmci_lagged_driver).

• Full PCMCI with conditioning sets is not yet bundled; see PC/FCI adapters.

API

• ts2net.causal.time_lagged.time_lagged_causality_network()

• ts2net.causal.pc.pc_algorithm()

• ts2net.causal.fci.fci_algorithm()