Print

Print
Hi - it's because we wanted the nonlinear interaction to be relative to the "baseline" of the minimum (i.e. "rest"), rather than based around the mean  - for various applications this makes the interaction more useful.   See more discussions wrt PPI analyses (e.g. on Jill's page) for more on this kind of thing.

Cheers


On 18 Apr 2013, at 06:24, Daniel CM <[log in to unmask]> wrote:

Hi all -

Can anyone point me toward a good explanation as to why FSL recommends setting the "minimum" and not the "mean" of component EVs to zero when setting up an interaction in FSL (or creating the interaction term manually, for that matter)?

Relatedly, is it sensible to interpret main effects (i.e. the effects of those component EVs), when an interaction effect is present - which would generally be interpretation of a standard regression result, if the main effect occurs in a) the same region as the effect of the interaction and b) a different region?

Cheers and all the best,
Dan






---------------------------------------------------------------------------
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
---------------------------------------------------------------------------