On 5/11/13 7:41 PM, Tudor Popescu
wrote:
Hi everyone, I resort to asking some questions here
again, after feeling increasingly stupid (maybe just tired) after
reading the wiki pages relevant for my analysis (ROI, thickness).
1) How should the contrast vector for a group comparison while
factoring out age and gender be built? I didn't quite understand
the example in the tutorial
based on intercept/offset. In my case, shouldn't I have 4
elements in the vector (group1, group2, age, gender), with the
first two elements receiving perhaps 1 and -1, since they are
directly compared, and the last two elements 0 and 0 since I am
not directly interested in their effect on the dependent variable?
Gender should be two groups (don't code it as a continuous
variable). This will give you 8 columns. g1, g2, male, female,
g1-age, g2-age, etc.
2) Are the structures described in the .label files produced by
mri_annotation2label in common (MNI) or in that subject's
individual space?
Individual space.
3) For my thickness analysis, I have in mind an ROI for which I
have a set of centre coordinates from a meta-analysis of relevant
functional studies. What is the best way of defining the mask for
this ROI as a function of those coordinates, rather than taking
the eponymous region from the Destrieux or the Desikan atlas?
Assuming that the coords are in mni152, convert it to mni305 (to do
this search for "coordinates" on the wiki). Then load up fsaverage
and find the vertex closest to this coordinate. The draw a label
around this vertex.
4) Exactly which of recon-all's outputs has to be inspected, for
each subject, in order to tell whether the surface reconstruction
has gone well?
Load up the surfaces over the nu.mgz or the T1.mgz and see if the
surfaces follow the anatomy properly.
5) Is it statistically correct to run the same design both with a
DODS FSGD and with a DOSS FSGD? It's not entirely clear to me what
assumptions each one (DODS, DOSS) makes of the data, and which one
I should select in my case.
The DODS allows for interactions between a continuous factor and a
discrete factor. Eg, the thickness-age slope could be different
between males and females, and you could detect this with DODS. With
DOSS, there is only one thickness-age slope for all groups, so there
is no way to detect an interaction.
doug
Thank you very much indeed.
Tudor
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer