Hi,
On 1 Sep 2006, at 04:44, Dan Wolf wrote:
> Hello,
> I am interested in finding regions where the task-related
> activation correlates across-subjects with
> another individual trait (in this case, an ERP amplitude measure).
> In general, I plan to take the first
> level individual analysis into a second level with the ERP
> amplitudes as an EV of interest. I have a
> couple of related questions:
> 1. Is it ok/recommended to use Flame1 and/or Flame2 for this
> purpose? I could do a simple
> Pearson's correlation at every voxel, or simple regression, but my
> hope/understanding is that
> Flame can take into account individual-level variance in a way that
> might make the analysis more
> statistically powerful and/or accurate - is that true for this kind
> of analysis?
Yes, we would recommend FLAME for this analysis, with the caveat of
course that you are only as flexible as the model you're fitting - so
if you just put in the ERP amplitude measure, the model assumes
linearity in the relationship between that and the BOLD response. You
can always also add in (for example) ERP^2 as a second covariate to
test for a second-order term in the relationship.
> 2. Assuming I am not interested in assessing the effect of the
> covariate on the measurement of the
> group mean, is there any reason to include a group mean EV (all
> 1s)? Or can I just do the analysis
> with a single EV, consisting of demeaned ERP amplitude measures for
> each subject?
Not at second level, no, as the data is not demeaned at second-level
(because it's often of interest). You still need a group mean EV,
which you will then ignore when you setup the contrasts.
Cheers, Steve.
------------------------------------------------------------------------
---
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
------------------------------------------------------------------------
---
|