Hi - this depends on exactly _what_ question you intend to be asking with
the ANOVA and with the paired-t-test analyses. In general, as they are
asking quite different questions, with different specificities and
sensitivites (e.g. different modelling power etc) there is no reason why
you should _have_ to mask in the way you are describing. If the t-test is
asking the right question and seems the right model for your design then
you don't need to further mask with the ANOVA results in general. Hope
this makes sense!
Cheers, Steve.
On Fri, 23 Jan 2004, Antti Korvenoja wrote:
> Dear all,
>
> I would like to present a question concerning proceeding from ANOVA to
> pairwise t-testing.
>
> We have performed analysis analogous to the ANOVA 1 factor 4-levels example
> in FEAT manual (in our case 3 levels) and in the resulting F-statistic
> certain areas survive the significance threshold used.
>
> In an analysis analogous to the pairwise t-test example in FEAT manual we
> see that in one of the comparisons between two levels there are voxels that
> survive the thresholding and overlap those seen in ANOVA. In addiotion there
> are activated voxels outside ANOVA areas.
>
> Should we use the ANOVA thresholded activation as mask for pairwise
> comparisons using t-test (i.e. should we discard voxels from further
> analysis that showed no effect in ANOVA), and if so, how to correctly set up
> that?
>
> Yours,
>
> Antti Korvenoja
> --
> Antti Korvenoja
> Biomedicum Helsinki, Radiology
> P.O. Box 700, 00029 HUS
> Finland
>
> tel: +358-9-47171786 : mobile: +358-40-7498469
>
Stephen M. Smith DPhil
Associate Director, FMRIB and Analysis Research Coordinator
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|