 |
 |
Hi Bing, There isn't a specific tool. However, the formula is somewhat simple so it should be straightforward to do it in, e.g., Octave or Python. 1) Start by computing a measure of the autocorrelation. Say you have timeseries for voxels X and Y stored in variables similarly named in Octave/Matlab. Then the correlation between two consecutive timepoints of X can be computed as: r_X = X(2:end)'*X(1:end-1)/X/X'; 2) Repeat for Y, obtaining r_Y. 3) Use Bartlett's formula to compute the degrees of freedom: N*(1-r_X*r_Y)/(1+r_X*r_Y), where N is the number of observations. Having said that, if this is a between-subjects analysis, and the scans were acquired under the same conditions, we expect that all subjects will have the same effective degrees of freedom, such that the Bartlett correction is not necessary -- it can be safely skipped. All the best, Anderson On 11 August 2017 at 08:53, Bing Yu <[log in to unmask]> wrote: > Dear FSL experts： > > In my functional connectivity research, I have to calculate Bartlett > correction factor (Fox,PNAS,2005, 102, 9673–9678) to correct the degrees > of freedom in Fisher's transformation of Pearson correlation maps. > > I will be very appreciate if I got advises on how to calculate > Bartlett correction factor using FSL (or its plugins), or wether there > were better ways to correct the autocorelation of fMRI time series other > than Bartlett’s theory. > > Thanks! > Bing Yu >