Mscoherence

Hi all,

I am not sure about the behavior of matlab mscoherence. I have 26 digital signals (boat instruments, pressures) and I want to calculate the coherence among them.

The sampling frequeny is about 2 Hz, and the number of each series is of 1890 elements.
I used this procedure to calculate each mscoherence:

nfft = 256;
fsample = 2.12;
[coh,f] = mscohere(UnResultsOctober_run19(:,j), UnResultsOctober_run19(:,i), hamming(nfft), nfft/2, nfft,fsample);
 
The problem is that the first values of coh are always close to 1, and then slowly (say from 0.1-0.2 hz) drop to lower values. I thought this is due to the windwing (hanning) of the signals, is it true? If yes, how do I understand which are the relevant values to consider ? (so which range of frequencies to consider?)

If not, it means that I understood nothing and I would appreciate some help :)

I tried to repeat the computation with higher nfft, to play a bit with it to understand how it affects the results, but the situation is doesn't change a lot; and after all I don't have particular needs in terms of frequency resolution.

I hope that I've been clear in my explanation. I can send the data if needed (is a 1890 X 26 array)
Thanks

Dario