Dear frmi/nilearn experts,

I’m working with ABIDE preprocessed as dataset to build a classification model for ASD. I’m working with multi-site dataset so i’d like to apply Fisher z-transformation to my connectivity measure matrix.

When I perform np.arctanh(X), where X is the vector of feature obtained by nilearn.connectome.ConnectivityMeasure(kind=‘tangent’,vectorize=True, discard_diagonal=True).fit_transform(A), where A is the matrix of timeseries, I obtain ‘nan’ values.

I understand that ‘np.arctanh’ return nan values where input are not in [-1,1] range.

So my question is: how can I perform Fisher z-transformation in this case? Could I use np.arctanh to the correlation matrix before perform tangent method?

Any advice would be greatly appreciated.

Best,

Giacomo