 |
 |
Hi Charlie, Please see below: On 11 July 2014 19:05, charlie <[log in to unmask]> wrote: > A tbss_sym FA between group analysis showed significant L-R differences. > ok > > A FAST analysis of the two groups also showed significant differences in > white matter volume. > ok > > Shapiro-Wilk (normality) and Levene (homoskedacity) Tests on the FAST data > set of the (combined )groups were both highly non-significant. > I understand that these tests were applied to the global volume of WM, so this looks ok (although these tests should be applied to the residuals, I don't expect the results to be much different). > Therefore, is it appropriate to conclude that neither non-normality or > variance in white matter volume between groups contributed to the between > group FA differences? Nope, it doesn't follow, because global WM volume doesn't have much to do with water diffusion as assessed via FA. > In addition, it might also follow that because TBSS constrains FA analyses > to within a common mean core FA skeleton, between group FA differences due > to total white matter volume difference is unlikely. > Nope, this is also a non sequitur I'm afraid: if we came up with something like a "cortex-based spatial statistics" (say, a "CBSS") to do VBM, we'd still need to account for the stretches and shrinkages before projecting to the "cortical skeleton". FA is unrelated to global WM volume not because of the skeletonization, but because these two quantities refer to different things. > I would think ideally that volume (demeaned) should have been included as > a covariate in the original glm design... > I don't think this is necessary for FA. All the best, Anderson