You mean C15-C21? Yes, sounds right.

 

Sounds right too. Only note that differently than the previous, this F-test may be rank deficient. Consider skipping one of the t-tests (any), e.g., use C22-C27 instead.

All the best,

Anderson


________________________________________
Van: FSL - FMRIB's Software Library [[log in to unmask]] namens Anderson M. Winkler [[log in to unmask]]
Verzonden: zaterdag 13 december 2014 19:33
Aan: [log in to unmask]
Onderwerp: Re: [FSL] 2x7 ANOVA - interaction effect

Hi Iris,

This design has problems: EV1 is a linear combination of the subject-specific EVs, and in order to check the interaction, the EVs that code for the stimulus differences need to be split between the groups. The contrasts also don't seem right.

Please, see attached two possible ways of coding. They are equivalent to each other and you can choose either -- results should be the same. You may not need all these contrasts, but I included just so that their construction is clearly visible. You may want to pick the ones you wish for the F-test(s) that you'd like to do.

Use this design (either of them) with randomise, and use the "-e design.grp" option with contrasts C1-C21. For the remaining contrasts, include also the option "--permuteBlocks".

All the best,

Anderson

PS: Just in case, .ods files can be opened with LibreOffice (open-source spreadsheet program, among other things).
PS2: The list won't take the attachment because of the size (>200KB). I'll leave it do download at this link for a couple of days: https://dl.dropboxusercontent.com/u/2785709/outbox/designs_iris.ods

On 13 December 2014 at 19:29, Anderson M. Winkler <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Hi Iris,

This design has problems: EV1 is a linear combination of the subject-specific EVs, and in order to check the interaction, the EVs that code for the stimulus differences need to be split between the groups. The contrasts also don't seem right.

Please, see attached two possible ways of coding. They are equivalent to each other and you can choose either -- results should be the same. You may not need all these contrasts, but I included just so that their construction is clearly visible. You may want to pick the ones you wish for the F-test(s) that you'd like to do.

Use this design (either of them) with randomise, and use the "-e design.grp" option with contrasts C1-C21. For the remaining contrasts, include also the option "--permuteBlocks".

All the best,

Anderson

PS: Just in case, .ods files can be opened with LibreOffice (open-source spreadsheet program, among other things).


On 12 December 2014 at 23:41, Goossens Lies (PN) <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Dear Anderson,

Somehow the text got hidden in attach to the message.
Our question was the following:


Dear FSL experts,

We conducted an fMRI study with 2 groups (30 subjects each) viewing 7 different stimuli. We expect activity to gradually increase/decrease in certain areas in the brain. We would first like to establish in which regions a difference between the two groups across the stimuli can be found (group x stimulus interaction effects).

We set up the following design matrix and contrasts (+F-contrasts). We have based this matrix on previous threads in the archives and the FSL wiki. However, we came across different approaches and we were wondering if this is the correct way.
Please find attached the design matrix that we came up with. Did we use the correct set up and contrasts for the main effects and interaction effect?

Many thanks in advance for the help!

All the best,

Iris

From: "Anderson M. Winkler" <[log in to unmask]<mailto:[log in to unmask]>>
Reply-To: FSL - FMRIB's Software Library <[log in to unmask]<mailto:[log in to unmask]>>
Date: vrijdag 12 december 2014 23:25
To: "[log in to unmask]<mailto:[log in to unmask]>" <[log in to unmask]<mailto:[log in to unmask]>>
Subject: Re: [FSL] 2x7 ANOVA - interaction effect

Hi,
Did you mean to send a question? The body of the message came empty...
Anderson


On 12 December 2014 at 09:36, SUBSCRIBE FSL IrisL <[log in to unmask]<mailto:[log in to unmask]>> wrote: