Paul,
>
> > 2. what test would you recommend I use for the second level analysis if
> > I want to determine the areas of activation for each condition (a, b, c)
> > and the areas activated during c but not a or b?
>
> A one-sample t-test should do the trick for your first question. This will
> test the null hypothesis that the sample mean is zero (i.e. no difference
> between condition X and implicitly modelled rest).
>
> Your second question unfortunately can't be addressed statistically. You
> cannot assert the null hypothesis and identify areas are NOT significantly
> activated by conditions a or b. This is a statistical issue, rather than
an
> SPM limitation! At a descriptive level, within SPM you could use exclusive
> masking of condition c activations by conditions a & b. However I stress
> that the resulting SPM should be treated purely in a descriptive fashion,
> and this is not a statistically valid way to approach the data!
>
I may be missing something obvious here, as I don't work with fMRI myself,
but...
Wouldn't it be possible to produce a first level contrast of c against a & b
(-0.5 -0.5 1) for each subject which could then the used for a one-sample
t-test at the second level?
If this is blatantly wrong it would be useful if someone with a greater
understanding could explain it for me.
Cheers,
Mark
========================
Dr. Mark Daglish
Clinical Research Fellow, Psychopharmacology Unit, University of Bristol
Visiting Clinical Fellow, MRC Cyclotron Unit, Hammersmith Hospital, London
Tel: +44 (0) 117 - 925 3066
Fax: +44 (0) 117 - 927 7057
email: [log in to unmask]
========================
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|