Dear Davy,
2009/11/8 闫伟 <[log in to unmask]>:
> Dear spmers,
> I'm doing MEG group analysis now. I want to find the differences
> between Group 1(7subjects) and Group 2(7subjects), Statistical
> Analysis Design is the attachment 1. (specity 2nd-level, two-sample
> t-test )
> When doing Result, if I select T-contrast, what should I enter
> Contrast Weights , [1 0] \ [0 1] \ [1] \ [-1] \ [1 -1] \ [-1 1 ] or
> others? (attachment )
To find the difference between groups you should enter either [1 -1]
(group1 > group2)
or [-1 1] (group2>group1).
> If I select F-contrast, what should I enter Contrast Weights , [1 0] \
> [0 1] \ [1] \ [-1] \ [1 -1] \ [-1 1 ] or others?
>
You should enter either [1 -1] or [-1 1] the results will be the same.
> When selected contrast weights [1 0](attachment 3 or 5), the result is
> the same as [1].
If you enter less coefficients than there are columns in your design
matrix, your contrast vector will be padded with zeros so [1] and [1
0] are different ways to specify exactly the same thing.
> While selecting [1 -1] or [-1 1](attachment 4 or 6),
> there was always no significance. Is it really no significance in
> fact, or I did something wrong?
Whole-brain FWE corrected test is quite conservative and unless there
are very strong and consistent effects they will not be significant.
One thing you can do is small volume correction for your areas of
interest DEFINED IN ADVANCE. But to get an idea what's happening you
can look at uncorrected results and see how close you are to getting
something significant.
> I'm a new face to the SPM and really
> don't kown 1\0\-1 means.
These are just coefficients that define a linear combination of the
columns of the design matrix that you are testing. In your case the
first column is group1, the second column us group2. If one of the
coefficients is 0 it means that the corresponding group is not
involved at all in the contrast so you are basically doing
single-sample t-test for the other group. In M/EEG source
reconstruction case this kind of test is not meaningful because the
images you get are always positive.
> When I want to find the result GROUP 1 and GROUP 1, Contrast weight [1
> 0], in fact its must be no significance, but its had significance in
> attachment 7 by SPM8b. I don't know why?
As I said this test is not valid, but I'd expect something to show up
in the contrast so perhaps you are doing something wrong. Try looking
at uncorrected results. In general there might be differences between
results of source reconstruction in different version of SPM because
the algorithm is still under development.
Best,
Vladimir
|