Is Granger or Coherence analysis better to use? And what programs would you suggest for one or the other? Are they equally publishable?

What type of data are you analyzing (e.g., fmri, eeg)?

The Granger causality analysis estimates the directionality of the modulation between two areas, while the coherence analysis does not estimate such a directionality. In other words, Granger causality analysis estimates the causal modulation, but the coherence analysis estimates the functional correlation.

The most popular implementation of the Granger causality is based on the stationarity assumption of the time series using an auto-regressive (AR) model. Suppose we have two time series at region A and region B. One AR model can be estimated on the time series at region A, and another AR model can be estimated on the joint time series at region A and region B. If the joint model can achieve a better prediction quantified by a statistically significantly smaller residual variance at the latter AR model compared to the former one, we consider that region B is Granger causing region A. Such an estimation process is not symmetric: we have to separately test if A is causing B and if B is causing A.

The detail of the AR model estimation, including model order selection and AR model parameter estimation, has been widely studied. There are quite a few publicly accessible computational resources. For example, ARFIT package is one offering the MATLAB implementation of the multivariate AR model estimation.

I hope this can be helpful.

Can you point to a couple of the 'publicly accessible computational resources' that you know of, so that we can try to get them included in NITRC directly?

Thanks.