Hi,
> thank you for your answer and sorry for the delay but I had some problems.
>>> I cannot say for sure, as I do not know how you've setup your EVs. If you have correctly setup EVs for the three groups (EV1-EV3) and for the covariates (EV4-EV5) then the contrasts look fine to me.
>
> The *.mat file is something like this:
>
> /NumWaves 5
> /NumPoints 60
> /PPheights 1.000000e+00 1.000000e+00 1.000000e+00
>
> /Matrix
> 1.000000E+00 0.000000E+00 0.000000E+00 1.000000E+00 4.400000E+01
> 1.000000E+00 0.000000E+00 0.000000E+00 0.000000E+00 3.500000E+01
> 1.000000E+00 0.000000E+00 0.000000E+00 1.000000E+00 2.800000E+01
> 1.000000E+00 0.000000E+00 0.000000E+00 1.000000E+00 3.800000E+01
> 1.000000E+00 0.000000E+00 0.000000E+00 1.000000E+00 2.400000E+01
> 1.000000E+00 0.000000E+00 0.000000E+00 1.000000E+00 3.400000E+01
> 1.000000E+00 0.000000E+00 0.000000E+00 1.000000E+00 2.800000E+01
> 1.000000E+00 0.000000E+00 0.000000E+00 1.000000E+00 3.800000E+01
> …
It's still hard to know what the first 3 EVs really are from this snippet, but I assume that they represent three groups (with ones for the subjects in that group and zeros elsewhere). However, you should demean the last two columns (separately).
>>> Each test is an independent statistical test in the GLM. However, if you want the equivalent of the ANOVA-style analysis then you would normally look at the F-test result to find which voxels are declared significant and then use the t-test results as post-hoc tests.
>
> Do you mean I've to take the significant results of the F analysis and use them as a mask (1 where significant and 0 differently) for the skeletons in the t-tests?
Yes - if you take the t-contrasts and use the F-contrast for contrast masking then this will achieve the desired result of showing you statistically significant results (based on the F-contrast) but then only showing the parts that respond to the specific t-tests.
>>> This is possible because the t-tests are one-sided tests while the F-test is two-sided and hence the threshold for significance is more stringent in the F-test (for the same total p-value).
>
> Ok thanks
All the best,
Mark
|