I similarly have a resting-state data set with two separate runs for the
same group of subjects. I have a couple questions based upon these posts.
How would I specify the contrasts to look for differences in the amount of
variance of the associated time-courses? This is just the concatenated data
from time1 contrasted to the concatenated data for time2 (for example)?
What would the GLM be? ...time1 and time2 concatenated together? Also, does
anybody have any recommendations on how to look for spatial map differences
in the new multi-session melodic analysis? ...I didn't realize the spatial
maps were "effectively fixed" in a multi-session concatenated or TICA
analysis. Thanks for your help.
On Aug 16 2007, Vitaly Napadow wrote:
>thank you, Christian. I think I understand - for my purposes, to test for
>spatial (not tempopral) differences, I will continue with my past
>On Thu, 16 Aug 2007, Christian Beckmann wrote:
>> On 15 Aug 2007, at 17:20, Vitaly Napadow wrote:
>>> hi all
>>> firstly, thanks for finally pushing this out. i think we were all
>>> chomping at the bit looking forward to the new release and it feels
>>> great to finally play with the new tools.
>>> i had completed an analysis of resting state data but want to go back
>>> and try the new tools now. i have 15 subjects and two separate rest
>>> runs that i am comparing from each subject. i assume i want to use
>>> Multi-session temporal concatenation to contrast the two rest runs.
>> Yes, temporal concatenation is probably what you want, though you still
>> will need to decide on what exactly you want to compare between runs,
>> e.g. the amount of variance in the associated time course (as a measure
>> of volatility) or any other quantity you could come up with such as mean
>>> so as not to compare apples to oranges, i want to contrast specific
>>> resting state networks. my question is, if i specify and 2nd level
>>> design.mat and design.con file with a 1 -1 contrast how will i know
>>> which resting state networks (evident on group maps) are associated
>>> with or correspond to which 2nd level output components?
>> Don't quite understand the question. In the new release version you
>> have the option of testing (in the GLM sense) each associated time
>> course and each associated session/subject mode using design and
>> contrast matrices. The GLm fit will be performed separately for each
>> time course and/or subject session mode vector.
>>> alternatievly, if i already did single-subject analyses and have mixing
>>> matrices etc with components corresponding to known resting networks and
>>> which can be pulled out to input into a 2nd level group analysis, are my
>>> options to contrast on a 2nd level the same as before?
>> fundamentally yes, the main difference being that in the case of
>> separate analyses you're bound to have differences in e.g. the
>> default-mode network between subjects. In the case of temporal
>> concatenation or full TICA you effectively fix the spatial maps to be
>> identical and only allow the time courses to be different. hope this
>> helps Christian
>> Christian F. Beckmann
>> University Research Lecturer
>> Oxford University Centre for Functional MRI of the Brain (FMRIB)
>> John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
>> [log in to unmask] http://www.fmrib.ox.ac.uk/~beckmann
>> tel: +44 1865 222551 fax: +44 1865 222717