Hi - I'm afraid that from your description I can't tell in your
design what are the subject IDs and what are repeat scans within
subjects etc....could you clarify what your design is?
Cheers, Steve.
On 18 Jul 2007, at 18:05, James N. Porter wrote:
> Howdy FSLers-
>
> I have a question about proper design matrix procedures.
>
> First, say I've performed lower-level FEATs and obtained 8 copes (A-
> H) for my 8 task conditions for each subjects' repeated scans.
> Next, I run a higher-level analysis that obtains the subjects'
> means for the 8 copes across their scans. Then I run a higher-level
> analysis to obtain contrasts between these 8 copes. My inputs are
> cope images, of which each subject has 8. Here is my EV matrix for
> this set up (for simplicity, as if I only had 2 subjects):
>
> A B C D E F G H
> cope1.feat/stats/cope1.nii 1 0 0 0 0 0 0 0
> cope1.feat/stats/cope2.nii 1 0 0 0 0 0 0 0
> cope2.feat/stats/cope1.nii 0 1 0 0 0 0 0 0
> cope2.feat/stats/cope2.nii 0 1 0 0 0 0 0 0
> cope3.feat/stats/cope1.nii 0 0 1 0 0 0 0 0
> cope3.feat/stats/cope2.nii 0 0 1 0 0 0 0 0
> cope4.feat/stats/cope1.nii 0 0 0 1 0 0 0 0
> cope4.feat/stats/cope2.nii 0 0 0 1 0 0 0 0
> cope5.feat/stats/cope1.nii 0 0 0 0 1 0 0 0
> cope5.feat/stats/cope2.nii 0 0 0 0 1 0 0 0
> cope6.feat/stats/cope1.nii 0 0 0 0 0 1 0 0
> cope6.feat/stats/cope2.nii 0 0 0 0 0 1 0 0
> cope7.feat/stats/cope1.nii 0 0 0 0 0 0 1 0
> cope7.feat/stats/cope2.nii 0 0 0 0 0 0 1 0
> cope8.feat/stats/cope1.nii 0 0 0 0 0 0 0 1
> cope8.feat/stats/cope2.nii 0 0 0 0 0 0 0 1
>
> Now say I wanted to know the results of the contrasts
> 1-- ((A-B)-(C-D)) - ((E-F)-(G-H))
> 2-- ((A-C)-(B-D)) - ((E-G)-(F-H))
>
> Would it be mathematically/statistically correct to simply perform
> a distribution of the signs/operators and end up with contrasts
> coded as:
> A B C D E F G H
> 1-- 1 -1 -1 1 -1 1 1 -1
> 2-- 1 1 -1 1 -1 1 1 -1
>
> Under this method, I can imagine two conceptually distinct
> contrasts that multiply out to have the same ±1 values, which I
> also imagine would be a bad thing. If so, would my analysis path be
> required to follow the old PEMDAS rules and find the values of each
> parenthetical contrast in the prescribed order? In other words, is
> it best to
>
> 1) define the elements of a higher-order contrast in the lower-
> level analyses, obtain subject means on the elements in a higher-
> level analysis, then contrast those results in one shot in a group
> means higher-level analysis, or
>
> 2) set up all desired higher-order contrasts from the get-go at the
> single-subject level, then simply obtain subject and group means
> for those4 copes (performing no extra contrasting) with the higher-
> level analyses?
>
> Thanks for all your help,
>
> -- Jim Porter TRiCAM Lab Coordinator Elliott Hall N437 612.624.3892
> www.psych.umn.edu/research/tricam
------------------------------------------------------------------------
---
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
------------------------------------------------------------------------
---
|