We are using 3dFourier to filter out low frequency drift.
If we include the retrend opton on it, everythings works as expected.
If we don't retrend the data, then when we do a fim we get a lot of pixels whose correlation
is 0.
The same pixels in the retrended data have non-zero correlations.
Since fim removes a linear trend (my understanding) then the correlation coefficients should be
identical regardless of whether or not we retrend after FFT filtering.
But they are clearly not. We looked at histograms of the correlation subbrick and it differs
also depending on whether or not the analysis is performed on retrended data.
Lee