Hi - yes that makes sense - and for these reasons I would use all group maps in the dual-regression, even if you only intend to then utilise the outputs from a couple.
Cheers.


On 24 Apr 2014, at 14:50, Lückmann Helen (PSYCHOLOGY) <[log in to unmask]> wrote:

Dear FSL experts,

I have a question about the use of dual regression (on a FIXed data-set) when one is only interested in few (1-2) components.
Is it sensible to only use the spatial maps of these components as input or does it actually make sense to include a couple more meaningful components?

The rationale behind the question is the following:
If the multiple regression behind dual regression works with partial regression coefficients (and components are not completely spatially independent), I would assume that by including more components, the time-course that is attributed to my components of interest is somewhat cleaner, as time-course variance pertaining to one of the other components will be attributed to those other components and not (erroneously) to my components of interest.
Not sure if this makes sense - statistically and with respect to the workings of dual regression - so I am grateful for any feedback!

Cheers,
Helen


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

Stop the cultural destruction of Tibet