Hi Eugene, Thank you very much for your help. My time series are z values (and I used age and gender as covariates, which were demeaned), is it ok? When I calculated the ROI-to-brain voxel-wise negative FC of the thalamus as ROI by using -1 for the group mean (and 0 for the covariates) in the GLM setup, I obtained some significant correlations only with some regions in the white matter... I also tried to get the negative effect (mean) in the one sample t test by firstly multiplying the data by -1 (following the GLM - FslWiki tutorial for single group average - one sample t test, in order to obtain the negative effect), but I obtained again significant correlations only with white matter... What do you think? Thank you again for your help, Matteo 2017-02-04 21:30 GMT+01:00 Matteo Antonio <[log in to unmask]>: > Hi Eugene, > > Thank you very much for your help. > > My time series are z values (and I used age and gender as covariates, > which were demeaned), is it ok? > When I calculated the ROI-to-brain voxel-wise negative FC of the thalamus > as ROI by using -1 for the group mean (and 0 for the covariates) in the GLM > setup, I obtained some significant correlations only with some regions in > the white matter... I also tried to get the negative effect (mean) in the > one sample t test by firstly multiplying the data by -1 (I attached the > screenshot of the tutorial that I followed), but I obtained again > significant correlations only with white matter... What do you think? > > Thank you again for your help, > Matteo > > 2017-02-03 21:09 GMT+01:00 Eugene Duff <[log in to unmask]>: > >> Hi Matteo >> >> On 3 February 2017 at 18:17, Matteo Antonio <[log in to unmask]> wrote: >> >>> Dear FSL e-mail list, >>> >>> >>> I'm doing a whole-brain voxel-wise functional connectivity analyses with >>> FSL for resting state fMRI data and I'm trying to perform a one-sample t >>> test to detect the significantly positively and negatively connected >>> regions (correlated and anti-correlated) with one ROI in my sample. >>> >>> >>> >>> When I did the one sample t-test for detecting the significantly >>> positively connected regions for the ROI, I got consistent results (I put 1 >>> in the correspondent contrast to get the group mean in the GLM setup). >>> >>> >>> But I found strange results when I tried to get the negatively connected >>> regions with the same ROI... According to the FSLwiki tutorial, firstly I >>> multiplied the original data used to test the positive mean by -1, and then >>> run the randomise function with the same GLM setup that I used for >>> calculating the positive mean. I also tried to directly use -1 (insted of >>> 1) in the GLM setup to get the group negative mean. But in both cases the >>> results are very strange (and similar between each other) and seem to >>> be incorrect. >>> >>> >>> >>> I think I'm doing something wrong... Can someone help me? >>> >> >> >> Are your time series demeaned? This could cause strange results. Really, >> to test negative correlations, all you should need to do is change the >> contrast from 1 to -1. >> >> Eugene >> >> >> >>> >>> >>> Thank you so much! >>> >>> Matteo >>> >> >> >