Dear all,

Does it make sense to apply variance smoothing in FSLNets (–v HWHM mm in randomise), when doing the statistics in correlation network matrices? I read that variance smoothing is useful for studies of less than 20 subjects, but can this be applied only with dual regression stage 2 or also with correlation matrices?

Another question: In FSLNets, concerning variance normalisation in timeseries: when should I use 0: no normalisation, 1: normalise for the subject standard deviation or 2: normalise each timeseries by its standard deviation?

Thank you,

Best,

John