Dear Tianhao,
> Then if I asume there is no rest condition during the experiment,
> such as [high low high low...]. And I specify two columns in the model,
> Then the contrast [1 0] reprsent 'high-low' or 'high-rest'? How to
> decide the 'rest' here? What is the meaning of contrast [1 -1]?
In general you can say, if you have two conditions, you can name them as
'activation' and 'rest' or 'high' and 'low', you just need to model one of
them. So, in your example the contrast [1 0] is 'high-low'. In this design
you can not test against a 'rest' condition. This two condition designs can
only test one condition against the other one, so the second condition can
be 'rest' or a appropriate control condition like 'low'.
> Another question: I understanded that there are two ways to do
> multi_subjects analysis. The first is to only do first level analysis;
> The second is to two-levels analysis. Then what is the difference
> between the contrast [1 -1] in the first way and the contrast [1 -1]
> after finishing two-sample Test in the second.
The first level analysis have usually the problem, that you have very high
degrees of freedom and therefore very high t-values. This first level,
fixed-effects analysis do not account for the size of your investigated
population.
This is, what the second level analysis does. The second level analysis,
which is a complete new statistics, results in degrees of freedom and
significance level, which are appropriate for your size of population.
So, if you estimate the contrast [1 0] for each subject (if you have only
two conditions and modelled as mentioned above), you can use the
corresponding con*** images for a one-sample t-test. Here, you can use the
contrast [ 1 ], which will display all voxel, which are activated in all
subjects and the significance is according to the size of the investigated
group (at least 12 subjects or more).
The interpretation of the contrasts in the first and second level differ in
the point of view. The first level says, these voxels are significant
according to this measurement of the subjects (with #### df's), the second
level says, these voxels are significant with respect to the size of
population, which is a more strictly result.
Good luck
Karsten
----------------------------------
Dipl. Phys. Karsten Specht
Medizin Center Bonn
Spessartstrasse 9
53119 Bonn
Germany
Phone: ++49-(0)228/90 81-178
Fax: ++49-(0)228/90 81-190
E-Mail: [log in to unmask]
WWW: http://www.mcbonn.de/Praxis/praxis15/fmri1.htm
|