 |
 |
Hi Alena, You can do the following: 1) Make a design matrix that has only the nuisance variables of your original test. 2) Run the GLM with your original voxelwise data and the design matrix from (1) using the command fsl_glm. Save the residuals. 3) Take the significant clusters from your randomise TBSS analysis and make a binary mask using fslmaths. 4) Apply the mask produced in (3) to the residuals produced in (2) using the command fslmeants, obtaining nuisance-free values within the significant regions for each subject. 5) Use your favorite graphing software to plot the variables of interest vs. the values from (4). Note, however, that the average across the voxels used for the figure isn't quite the same as the voxelwise information used in the original randomise test. It may be helpful for visualisation but no new conclusions should be taken and, crucially, no new p-values, as this would be circular with the initial analysis. All the best, Anderson On 12 April 2018 at 03:57, Alena Piovar <[log in to unmask]> wrote: > Dear FSL experts, > > I have a question regarding correlation analysis. I applied design matrix > to examine relationship between WM and clinical variable in TBSS and found > significant clusters. Then, I calculated correlation coefficients and > betas for the significant correlations that I have. > > Can you advise me how to calculate the each subjects values to create a > graphical presentations of the outcomes? > > Best regards > > Alena > > >