FIAR (version 0.6)

diffGranger: Directed conditional granger causality

Description

Compute the difference conditional granger causality of multivariate timeseries.

Usage

diffGranger(data, nx = 1, ny = 1, order=1, perm = FALSE, bs = 100)

Arguments

data

object containing all observations (rows) and variables (columns) that are being considered. The variables should be ordered as follows: First the variables that are supposed to granger cause a set of other variables (>=1). Then the set of variables (>=1) that are Granger caused by the first set of variables. Finally, a set of variables to condition on(>=1).

nx

The number of variables (>=1) that Granger cause a set of other variables (default = 1), conditioned on a third set of variables (>=1).

ny

The number of variables (>=1) that are Granger caused by the first nx variables (default = 1), conditioned on a third set of variables (>=1).

order

Autoregressive order (>=1) of timeseries. Can be computed using ARorder().

perm

Logical. If perm = FALSE (default), only the Granger causality measure is produced. If perm = TRUE, the Granger test is computed and a permutation test is performed to do inference.

bs

Number of permutation samples. Only works when perm = TRUE. Default=100

Value

Partial Granger causality measure F1 plus p -value (Only when perm=TRUE).

Details

The total linear dependence between X and Y can be divided in three components: a directed influence from X to Y, a directed influence from Y to X and an undirected instantaneous influence between them. The difference granger causality from X to Y computed in the function diff.granger() subtracts the conditional granger causality from Y to X from the conditional granger causality from X to Y. This can be used as a measure of how much stronger (weaker) one directed influence is compared to the opposite directed influence.

References

Roebroeck, A., Formisano, E., Goebel, R. (2005). Mapping directed influence over the brain using Granger causality. NeuroImage 230-242.

See Also

condGranger, pdiffGranger

Examples

# NOT RUN {
# Example data with 5 regions x, y, z, q, w
head(grangerdata)

# Calculate AR() order of the data
ARorder(grangerdata, max=10)
 
# Compute difference conditional granger causality of region x to regions y o
# and z, conditional on regions q and w
F <- diffGranger(grangerdata, nx=1, ny=2, order=3)

# Compute F and bootstrap H0 distribution
F <- diffGranger(grangerdata, nx=1, ny=2, order=3, perm=TRUE, bs=50)
# }