Dear Alexander,
It is quite possible for one of the constituent T-stats that form the F-test
to give more voxels above threshold. This is because the thresholds in
the two cases are somewhat different - since the T-test is one-sided it has
a slightly lower threshold. If the results are *very* different then this would
be a cause for concern, but if they just differ a little then that is fine and
to be expected.
There is a range of ways that people report results, from images to worded
descriptions of the anatomy to tables of cluster sizes and coordinates. It
really depends on how important the various parts are for your study, so if
the individual T-contrast is of great interest then report this separately, otherwise
just report the overall F-contrast results. Sorry, there's no universal answer here.
All the best,
Mark
On 3 Aug 2011, at 23:36, Alexander Olsen wrote:
> Yes, I guess I have to look into that. Now I have encountered a new problem I hope you can help me with. I have performed a repeated measures (1-2-3) analysis in FEAT and got a significant F-statistic. When I run the cluster command I find a signigifant cluster with a size of 1035 voxels. However, When I do the same for one of the simple contrasts (2 vs 4) which is the only significant contrast (apart from the F-stat). My first thought was that it was strange that the cluster from the simple contrast was larger than the cluster for the main effect in the same test... But this may not be strange at all? In addition I wonder how I should report results like this. Should I report both the main effect and various simple contrasts in a table, or should I just state that there was a significant main effect, and then report results from relevant simple contrasts? I know this may be basic knowledge for many of you, but I hope you can help me with some hints.
>
> All the best,
> Alexander
>
|