Hi Robert,
The Subject-mode vector is basically the vector of loadings on to the
associated Eigenvector, so yes, it does represent the fit wrt to the
rank-1 approximation. These plots are provided as an indicator for how
consistent effects are across the population assuming the same
temporal response.
Wrt the second question: you could perform spatial regression, i.e.
regress the DMN identified from separate ICA runs against a pre-
defined RSN mask (as in Greicius et al. Default-mode network activity
distinguishes Alzheimer's disease from healthy aging: evidence from
functional MRI. Proc Natl Acad Sci USA (2004) vol. 101 (13) pp.
4637-42). This then ignores the temporal characteristics altogether
and only looks across space (similarly, if you use GIFT you will
ignore the temporal characteristics if you perform stats on the back-
projected maps only).
The danger is that y
The alternative is to look across time and come up with summary stats
to compare, e.g. variance of the time courses or Hurst exponent or or
or, your call. You can try different ones but then don't forget to
correct for multiple comparisons ;) We are currently working on
evaluating various options and can hopefully give better advice in the
future.
hth
Christian
On 31 Jan 2008, at 20:41, Robert Kelly wrote:
> Dear Christian,
>
> Thank you for your quick reply to my questions and for reassuring me
> that
> the first Eigenvector timecourse plot produced for multi-session
> temporal
> concatenation is irrelevant for resting state analyses.
>
> Some follow-up questions: 1) Exactly what does the Sessions/Subject
> mode
> plot at the bottom of the component report page represent?
> Specifically,
> for resting state data (or any case where a common time course is not
> assumed to exist) does the plot represent the effect size of the fit
> for
> each session/subject with the component displayed or is it based on
> the
> rank-1 approximation? 2) Do you have any suggestions for how to
> test for
> group differences using some measure of how well each group fits
> with the
> given component?
>
> Thanks again,
> Robert
|