[FieldTrip] interactions between two factors

[Marco Buiatti]

Dear FieldTrippers,

some time ago I have posted the message below concerning how to
compute the statistical interaction between two factors in an EEG
study with the FieldTrip cluster-based statistical analysis. Since I
believe it is a problem of general interest, I was confident I would
have received some replies, no matter how critics. But I had no reply
and I am trying to guess why:

- the problem is trivial, I should go back to my statistics books and
solve it myself;
- the problem is ill-posed, I should go back to my statistics books
and reformulate it correctly;
- the problem is tabou, no one dares commits to a solution because it
could be a wrong one.
- the problem is solved: I should read message number #.

Thanks a lot for your feedback,

Best,

Marco

On 25 May 2012 15:58, Marco Buiatti <marco.buiatti at gmail.com> wrote:
> Dear FieldTrippers,
>
> I am analysing an EEG study with 2x4 factors: one varies between 4
> parametrically varying levels (1 to 4), the second between two levels.
>
> I have three questions concerning the use of Fieldtrip cluster-based
> non parametric statistical analysis in this case:
>
> 1) How to compute the interaction between the two factors. Let's start
> from the simplest case of a 2x2 design, factors varying between values
> A1 and A2 for the first factor, B1 and B2 for the second. Please tell
> me if it is correct to compute the interaction by:
> - computing the difference diffA=ERP(A1)-ERP(A2) separately in
> condition B1 and B2, for every subject
> - performing a within-subjects statistical analysis between diffA in
> condition B1 and diffA in condition B2 (function
> statfun_depsamplesT.m).
>
> 2) Now consider that factor A varies parametrically between values 1
> to 4. For the main effect of this factor, I have used the Fieldtrip
> function statfun_depsamplesregrT.m and I'm satisfied with it. Is it
> correct to compute the interaction by
> - computing the regression
> regrA=regression(ERP(A1),ERP(A2),ERP(A3),ERP(A4)) (computed as inside
> function statfun_depsamplesregrT.m) separately in condition B1 and B2,
> for every subject
> - performing a within-subjects statistical analysis between regrA in
> condition B1 and regrA in condition B2 (function
> statfun_depsamplesT.m)?
>
> 3) Since BEFORE looking at the data (this is to prevent Eric's
> contestation...) I expect a dipolar topography for the regression
> (data are in average reference), I would like to combine into a joint
> cluster negative and positive clusters. I have tried by changing
> statfun_depsamplesregrT.m by just taking the absolute value of the
> regression, but I get weird results (a huge, non significant cluster).
> Is it possible that since values are now all positive, I should use a
> different statistical test at the single bin level? Any other
> suggestions?
>
> Thanks in advance for your help,
>
> Marco
>
>
>
> --
> Marco Buiatti, PhD
>
> CEA/DSV/I2BM / NeuroSpin
> INSERM U992 - Cognitive Neuroimaging Unit
> Bât 145 - Point Courrier 156
> Gif sur Yvette F-91191  FRANCE
> Ph:  +33(0)169.08.65.21
> Fax: +33(0)169.08.79.73
> E-mail: marco.buiatti at gmail.com
> http://www.unicog.org/pm/pmwiki.php/Main/MarcoBuiatti
>
> ***********************************************



-- 
Marco Buiatti, PhD

CEA/DSV/I2BM / NeuroSpin
INSERM U992 - Cognitive Neuroimaging Unit
Bât 145 - Point Courrier 156
Gif sur Yvette F-91191  FRANCE
Ph:  +33(0)169.08.65.21
Fax: +33(0)169.08.79.73
E-mail: marco.buiatti at gmail.com
http://www.unicog.org/pm/pmwiki.php/Main/MarcoBuiatti

***********************************************