Here are some thoughts.
The reason you don't see options like you do in SPSS is that its hard to correct across space and contrasts. In SPSS, you are correcting across contrasts, but there is only 1 "voxel".
I would do the following:
(1) ANOVA with 5 groups, assess with F-test if there are any group differences.
(2) If you find group differences, then you can mask your data and do your post-hoc tests.
This will help control the error rate as you know that there is at least 1 group difference in the post-hoc tests.
The alternative of doing 5 separate tests doesn't get around the issue of multiple comparisons across comparisons.
There are people working on how to accurately do the joint correction across space and contrasts, but I'm not sure if anything has been released or published.
If you want to control across time and space, you could control across space, then use FDR or Tukey or another approach to control the tests across contrasts. This would require writing your own code and would like be confusing to the reader as each voxel would have a different threshold for being significant. For example, here are two voxels (each row represents a different comparison) corrected p-values:
| 0.001 |
0.001 |
| 0.001 |
0.025 |
| 0.025 |
0.025 |
| 0.045 |
0.045 |
| 0.049 |
0.049 |
If we use FDR, then the first voxel has 3 significant findings and the threshold would be 0.025, while the other voxel threshold is 0.001 and the 0.025 voxels wouldn't be significant.
Hope this helps.
