On 03/06/2014 10:35 AM, Tudor Popescu wrote:
Sorry, couple more questions:
- I know that QDEC's/mri_glmfit's equivalent in FSL ("randomise")
applies the GLM by doing permutations, and so already corrects for multiple comparisons in the process. How similar is this to what Monte-Carlo does (i.e. controlling FWE via repeated simulations) upon the initially uncorrected QDEC results?
By default QDEC users the monte carlo simulation. FSL randomise can do either monte carlo or permutaion. mri_glmfit-sim can also do permutation.
- In FSL, a comparison between groups A and B is tested by defining
both a "A>B" as well as a "B>A" contrast – the reason being (as far as I understand) that taking the logical complement of "A>B", i.e. inverting the t-values in the statistical map, is *not* equal to the map produced by the cotnrast "B>A". How come, then, that QDEC phrases the group contrast in a "two-tailed" way (i.e. "do groups A and B differ")? Is the dichotomy not necessary in QDEC, or is it simply done behind the scenes?
What do you mean by "inverting the t-values"? It is the same as changing the sign of the t values. In FS we preserve the sign in our "sig" maps (sign(t)*-log10(p)).
- I've read papers where the thickness from *native*-space was used
in the analyses, even though initially T1-weighted images were initially aligned to the ICBM 152 template. Why isn't thickness from this *standard*-space used, as happens in FSLVBM with grey matter density? Why even register to the template in the first place if you're then going to go back and use native-space values?
I'm not sure what you mean here. If someone warped their T1 data to some template space and then computed thickness, then that thickness will not represent the thickness from the individual and the group stats will be skewed by this doug
Thanks!
On 6 March 2014 10:20, Tudor Popescu <tudor3@gmail.com mailto:tudor3@gmail.com> wrote:
Thank you very much Doug. Tudor On 6 March 2014 03:53, Douglas Greve <greve@nmr.mgh.harvard.edu <mailto:greve@nmr.mgh.harvard.edu>> wrote: On 3/5/14 5:25 PM, Tudor Popescu wrote:Hello, I have some questions on doing group comparisons with thickness, area and volume. Many thanks in advance for any help! 1) For a DOSS design with group and gender as categorical factors, I see that an interaction contrast ("Is there a group-gender interaction in the mean thickness?") still exists - but what does this contrast mean, given that DOSS by definition doesn't allow for interactions?Are you using QDEC? If so, don't use the DOSS as the contrasts are incorrect. It is possible to have an interaction among the categorical factors with a DOSS.2) it makes sense that measures such as thickness are analysed vertex-wise in QDEC, however what does it mean when the dependent variable is area or volume - measures that do not make physical sense for a single vertex but only at the level of a region consisting of *several* vertices?The interpretation is a little more difficult. Each vertex is assigned an area equal to the average of the triangles adjacent to it. This is just a value that can be mapped to a common space like any other value (eg, thickness) (but there is a special jacobain correction to account for stretching or compression). Smoothing reduces the effect of having different sized triangles. One can think of it like this: in the common space (fsaverage) image having a patch of a certain size. When you mapped that patch back to each individual, how big would that patch be? You could then do group statistics on that number. In this way you could analyze the entire hemisphere. Now imagine doing this but making the patch smaller and smaller.3) For values extracted from atlas regions with aparcstats2table, it seems that the product of the extracted CT and area is in the same order of magnitude as the extracted volume, but never really the same or even close – why, when the volume of a region should theoretically be the product of its surface area by its thickness?It is an issue of how it is computed. Sum(CT*Area) != Sum(CT)*Sum(Area). When computing volume, CT*Area is computed for each vertex then summed across vertices. dougTudor _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer