Hi Steve,
Thanks for the response. It must be slightly more complicated than
demeaning since the result is not equivalent to the corrcoef() calculation
in matlab, and the relationship is more than the 1/n-1 factor. Presumably,
this is related to the other part of the denominator in the calculation of
the correl coefficient, related to the multiplication of variance terms
for each data set. Even if the data is demeaned, the calculation of sqrt
(a*a'+b*b') which I think fslcc is doing, won't be equivalent. Does that
sound right?
Thanks,
Gaby
|