Source coherence

Matsuhashi, Masao (NIH/NINDS) matsuham at NINDS.NIH.GOV
Wed Jun 1 04:20:30 CEST 2005

Hi Robert,

> I had a look at your data, and at the computations that you perfomed
> for the analysis. Your frequencies of interest are 4.6 and 5Hz, and
> your time windows are 10 seconds long (right?).


Based on your suggestions, I did some additional analyses with precaution
- cfg.taper='hanning' instead of the default 'dpss'
- Add coherence calculation at the frequency which shouldn't have meaningful
- Calculate and show coherence between MEG channels and Reference channels
using topoplotER
- Check the topological map of power distribution for possible artifacts

However, I still have the same problem.
I put a new data to
There are two reference signals, with frequencies of interest of 12 Hz and
15 Hz, respectively.
The coherence between MEG sensors and reference signal are around 0.8 - 0.9
at 12 Hz, but below 0.35 at 15 Hz. However the coherence value of
distributed source activities at 12 Hz was below 0.06, while it was above
0.1 at 15 Hz.

> coherence. But I still would not be too confident about it, since the
> influence of artifacts on the filter can be difficult to predict.

In the case above, 12 Hz reference signal may be too close to the dominant
frequency and can cause problem. However, when I calculate the coherence at
12 Hz between MEG and 15 Hz reference signal, the same thing happened again.
The coherence between EMG sensors and reference signal was below 0.3 as
expected, but source coherence peak was more than 0.1!? (Remember that the
coherence at 12 Hz with true reference signal was no more than 0.06)

Could you tell me what is going wrong, please?


Masao Matsuhashi, MD, PhD
Tel: 1-301-594-9139
E-mail: MatsuhaM at

More information about the fieldtrip mailing list