Thanks Jeanette.
I'm not sure if I really have correlations or not. From the original data, I used Feat to preprocess (motion correction, filtering, smoothing), entered a blank EV and added motion to model along with other physio regressors. From the resulting residuals (res4d), I extracted the timecourse of the seed using fslmeants. Then I performed another Feat analysis using the res4d file as input and the timecourse of the seed as an EV. I thought that the resulting zstat image shows voxels that have a correlated timecourse with that of the seed, but I'm not sure if the values in the voxels are 'r'.
Joyce
________________________________________
From: FSL - FMRIB's Software Library [[log in to unmask]] On Behalf Of Jeanette Mumford [[log in to unmask]]
Sent: October 27, 2011 10:28 PM
To: [log in to unmask]
Subject: Re: [FSL] fisher r to z for resting state fmri
Hi,
First, double check that the numbers you have really are correlations or the transform won't make any sense.
Then you can use fslmaths using the equation for the fisher's z transform
http://en.wikipedia.org/wiki/Fisher_transformation
1) Calculate 1+corr
2)calculate 1-corr
3) calculate ln ((1+corr)/(1-corr)) and multiply this result by 1/2.
Then you either need to trick feat into using the new FZT'd image you created or you ran use randomise to run the group analysis.
Cheers,
Jeanette
On Thu, Oct 27, 2011 at 5:02 PM, Joyce Chen <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Hello,
I am wondering how I can perform a Fisher's R to Z transform for my resting state data in FSL. For each subject, I ran Feat to look for whole-brain correlations with my timecourse of interest. How do I convert these maps of correlation coefficients to a map of Gaussian distributed values, so that I can perform higher-level analyses?
Thank you for your time and assistance.
Regards,
Joyce
|