On Tue, Nov 27, 2012 at 4:42 AM, Judith Suttrup <[log in to unmask]> wrote:
> Dear SPM list,
>
> We have a simple 2x2 factorial design with two groups and two tasks. We expect the following activation pattern for our regions of interest: group1_task1 > group2_task1 > group1_task2 > group2_task2 (activation increase for task2 -> task1 is identical between groups). To test for our hypothesized activation pattern, we run a full factorial design and assigned the following weights to our predictors: [1 0.3 -0.3 -1] (assuming a linear relationship as no other prior assumptions can be made). Is this a valid analysis? Can we interpret the statistical significant areas obtained by this analysis as regions that are enhanced activated in group1 for a process that is more relevant in task1 than task2?
Here are a couple of comments:
(1) You should be using a flexible factorial model with a subject
factor OR GLM Flex.
(2) The null hypothesis doesn't match the contrast.
(3) The null hypothesis can't actually be tested as you can't
statistically say two things are identical.
(4) What I think you want to ask is whether t1-t2 is greater in one
group compared to the other. If this is the case, then create the
flexible factorial model and use the contrast 0 0 0 0 1 -1 -1 1
zeros(1,n). This will identify areas where the task difference is
different between groups.
(5) Your existing contrast really doesn't inform you about anything.
> Any advice is appreciated!
>
> Kind regards,
> Judith Suttrup
|