Dear Christian,

Thank you very much for your very helpful response. I have been using your toolbox and I find it excellent (thanks!). I have not realy used the stats section because I am working withe large samples (~400 scans) and copying files to create groups in different folders is time consuming and error prone. Instead I prefer learning well and using multiple regression for all my analyses even when ANOVA could do the job. Can you see a problem with this approach? I am using spm2, is file management different in spm5?

Thanks again for your help and to the mailing list in general for this great resource.

Nic


-----Original Message-----
From:	Christian Gaser [mailto:[log in to unmask]]
Sent:	Fri 22/06/2007 19:22
To:	[log in to unmask]; Nicolas Cherbuin
Cc:	Christian Gaser
Subject:	Re: interactions with continuous variables (VBM with SPM2) 2nd attempt

Dear Nicolas,

I had the same problems and tested different models for interaction including your idea of 
multiplying the covariate with the group coding. However, this will not give correct parameter 
estimates and the right design matrix is:

GroupA GroupB CovariateCA CovariateCB 
1		0		34                 0
1		0		22                 0
1		0		25                 0
0		1		0                   15
0		1		0                   29
etc...

Then you can test the interaction with the contrasts "0 0 1 -1" and "0 0 -1 1". To model the 
interaction (its not very straightforward) you may try the vbm2 toolbox
http://dbm.neuro.uni-jena.de/vbm

There is an option for statistical analysis. Simply select "cross-sectional design", define the 2 
groups and the covariate and select "interaction with group" (no centering).

Best,

Christian

-- 
____________________________________________________________________________

Christian Gaser, Ph.D.
Assistant Professor of Computational Neuroscience
Department of Psychiatry
Friedrich-Schiller-University of Jena
Philosophenweg 3, D-07743 Jena, Germany
Tel: ++49-3641-935805	Fax:   ++49-3641-935280
e-mail: [log in to unmask]
http://dbm.neuro.uni-jena.de

On Fri, 22 Jun 2007 16:20:13 +1000, Nicolas Cherbuin <[log in to unmask]> wrote:

>Sorry for the repeat post but I did not get any answer.
>
>
>
>Dear SPMers,
>
>I have spent the last couple of hours searching the list for information
>on interactions with continuous variables in vbm (SPM2) but I am still
>very confused.
>
>I have two groups (A and B) and a covariate C. I would like to test the
>interaction between groups and the continuous variable C but I am not
>sure about the correct way to go about it. 
>
>I have set up the following variables
>
>GroupA GroupB CovariateC 
>1		0		34
>1		0		22
>0		1		15
>1		0		25
>0		1		29
>Etc..
>
>I then was not sure whether to test the interaction between groups and
>covariates I should simply multiply groupA by covariateC to produce
>AxCInteraction and then use the following contrast
>
> GroupA GroupB CovariateC AxCInteraction
>0		0		0 		1
>
>Or whether I should set up two interaction variables for each group
>variable and test the interaction with the following contrast
>
>GroupA GroupB CovariateC AxCInteraction BxCInteraction
>0		0		0 		1		-1
>
>Your assistance would be greatly appreciated
>
>Nic
>===========================================================
=============