Hi - for a given contrast, where the model includes several EVs, you
don't need to correct for multiple EVs. The number of EVs will affect
the degrees of freedom of the modelling and this is taken into
account automatically by the various GLM programs in FSL.
However, what your colleague may be referring to is if you are
running several different contrasts (in this case it sounds like
maybe you just have a simple [1] contrast for each EV you're testing,
one at a time) - in this case, then strictly, if the different tests
are independent of each other, and you are just looking for any
effect in any of the contrasts with no particular prior hypothesis
about what you're looking for, then yes you should correct the stats
for multiple contrasts (eg with Bonferroni). The software doesn't
automatically do that for you; most people in imaging conveniently
don't worry about this.
Cheers, Steve.
On 6 Jul 2007, at 00:34, Ping-Hong Yeh wrote:
> Hi ALL,
>
> I ran GLM randomise several times to test voxelwise significance
> for each
> EV that has been demeaned and covaried with one confounder. While
> evaluating
> the output file, maxc_ttest, which has been corrected for multiple
> comparsions using suprthreshold cluster test, one of my colleagues
> argured
> that correction for multple EVs are still required.
>
> Any comment on this?
>
> Best,
>
> Ping-Hong Yeh
------------------------------------------------------------------------
---
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
------------------------------------------------------------------------
---
|