 |
 |
Hi Jeremy, The original suggestion you made (Data=AD, EV1=column of 1's, EV2=voxelwise RD) is fine to calculate this sort of voxelwise correlation. All the best, Mark On 22 Nov 2013, at 18:48, Jeremy Strain <[log in to unmask]> wrote: > Dear Dr. Smith, > I wish it were just that. Here is the situation in more detail. In our paper we had patient longitudinal data showing Timepoint 1 had increased AD RD and MD but not FA compared to controls. We surmised that AD and RD proportionally increased which is the reason FA was not significant but all the three diffusivity measures were. The reviewer felt this was not an acceptable interpretation based on this evidence but that we could support our claim if we showed that at the voxel level and I quote "This might be only true if the proportional AD and RD value change occurs within the same voxel. Using voxel-based AD-RD correlational maps may clarify this point." Thus the reason for my original question. > > Thank you, > Jeremy > > > --------------------------------------- > Hi - I’m guess that the reviewer intended you to correlate the axial/radial diffusivity against the same set of subject regressors as originally done with FA - not against each other? So you can use tbss_non_FA (see wiki) for that? > > Cheers. > > > > On 21 Nov 2013, at 05:10, Jeremy Strain <[log in to unmask]> wrote: > > Hello, > A reviewer disagreed with one of our conclusions and suggested we could prove our claim by creating Axial Diffusivity - Radial Diffusivity voxel based correlational maps. I wanted to verify that this can be accomplished by using the AD data as an input and loading the RD data as an additional voxel-dependent EV and setting up the contrast as normal. > > Hypothetically no other variables: > Design: > EV1 > 1 > 1 > 1 > 1 EV2 = RD data > ... > > Contrasts: > > 0 1 --> AD and RD positively correlate > 0 -1 --> AD and RD negatively correlate > > Thanks, > Jeremy