FieldTrip

Note that this reference documentation is identical to the help that is displayed in Matlab when you type “help ft_connectivity_granger”.

  FT_CONNECTIVITY_GRANGER computes spectrally resolved granger causality.
  
  Use as
    [GRANGER, V, N] = FT_CONNECTIVITY_GRANGER(H, Z, S, key1, value1, ...)
 
  where 
    H is the spectral transfer matrix, Nrpt x Nchan x Nchan x Nfreq (x Ntime),
       or Nrpt x Nchancmb x Nfreq (x Ntime). Nrpt can be 1.
    Z is the covariance matrix of the noise, Nrpt x Nchan x Nchan (x Ntime),
       or Nrpt x Nchancmb (x Ntime).
    S is the cross-spectral density matrix, same dimensionality as H
 
  additional options need to be specified as key-value pairs and are:
    'dimord'  = required string specifying how to interpret the input data
                supported values are 'rpt_chan_chan_freq(_time) and
                'rpt_chan_freq(_time), 'rpt_pos_pos_XXX' and 'rpt_pos_XXX'
    'method'  = 'granger' (default), or 'instantaneous', or 'total'.
    'hasjack' = 0 (default) is a boolean specifying whether the input
                contains leave-one-outs, required for correct variance
                estimate
    'powindx' = is a variable determining the exact computation, see below
 
  If the inputdata is such that the channel-pairs are linearly indexed,
  granger causality is computed per quadruplet of consecutive entries,
  where the convention is as follows: 
 
   H(:, (k-1)*4 + 1, :, :, :) -> 'chan1-chan1' 
   H(:, (k-1)*4 + 2, :, :, :) -> 'chan1->chan2'
   H(:, (k-1)*4 + 3, :, :, :) -> 'chan2->chan1'
   H(:, (k-1)*4 + 4, :, :, :) -> 'chan2->chan2'
 
  The same holds for the Z and S matrices.
 
  Pairwise block-granger causality can be computed when the inputdata has
  dimensionality Nchan x Nchan. In that case powindx should be specified,
  as a 1x2 cell-array indexing the individual channels that go into each
  'block'.