Enumerate bootstrap paths between two variables in a VARMA-LiNGAM model

Builds the time-expanded graph for every bootstrap sample and enumerates all directed paths from the source (at from_lag) to the destination (at to_lag), reporting each path's bootstrap probability and median effect. Port of the Python reference VARMABootstrapResult.get_paths.

Usage

get_varma_paths(
  result,
  from_index,
  to_index,
  from_lag = 0,
  to_lag = 0,
  min_causal_effect = NULL
)

Arguments

result: a VARMABootstrapResult object
from_index: source variable (1-based)
to_index: destination variable (1-based)
from_lag: lag of the source (default 0); must not exceed the AR order p
to_lag: lag of the destination (default 0); must satisfy to_lag <= from_lag
min_causal_effect: minimum |effect| threshold (NULL = 0)

Value

a data frame (path, effect, probability), one row per distinct path

Details

Node indices in the returned path are 1-based positions in the time-expanded graph: column j of block L (lag L) corresponds to index n_features * L + j.

Only the instantaneous and AR (psi) blocks enter the time-expanded graph; the MA (omega) blocks describe effects of past unobserved disturbances, which are not nodes of the variable graph (the Python reference does the same).

Examples

s <- generate_varmalingam_sample(n = 300, seed = 42)
bs <- lingam_varma_bootstrap(s$data,
  n_sampling = 5L, order = c(1, 1), criterion = NULL,
  reg_method = "ols", prune = FALSE, seed = 1, verbose = FALSE
)
get_varma_paths(bs, from_index = 1, to_index = 3)
#>      path      effect probability
#> 1 1, 2, 3 -0.26152591           1
#> 2    1, 3 -0.09018331           1