 |
 |
Hi Joy, Yes, that's right. More interpretation for each contrast are possible, though: the [1 -1 0 0 ...]' can mean faster decrease in patients, but also slower increase in patients. It can also mean decrease in patients and increase in controls. Likewise, [ -1 1 0 0 ...]' can mean faster decrease in controls, but also slower increase in controls. It can also mean decrease in controls and increase in patients. All the best, Anderson On Friday, December 20, 2013, Joy Matsui wrote: > Hi Anderson, > > I have very delayed follow-up questions. > > I was a bit confused by the example to compute a difference image here ( > http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#Randomise_details-11) because > it's subtracting the second image from the first image, but then I thought > maybe what really matters is how you set up your contrast matrix... > > That being said, I just wanted make sure I have my logic correct when > finding the difference image and evaluating change in disease groups versus > controls. If my hypothesis is "FA decreases faster in the disease group > than the control group" and if I subtract my first time point images from > my second time point images (using the command: fslmaths FA_time_2 -sub > FA_time_1 FA_diff), then my contrast should be "1 -1 0 0…" where the first > column is controls and the second column is the disease group. If my FA at > time point 2 is indeed lower than time point 1 and the disease group has a > faster decreasing FA, then my difference for the disease group will be more > negative than the difference for controls and a "1 -1 0 0…" contrast would > test for that. And if I'm testing to see if a DTI scalar is increasing > faster in the disease group than controls, then I'd reverse my contrast. Am > I thinking about this correctly? > > > Thank you, > > Joy > >