 |
 |
Dear Alexandre, You can use spm_orth.m to orthogonalise your regressors. This implements Gramm-Schmidt orthogonalisation where the linear part explained by your first regressor is removed from your second regressor. Regarding your second question, using two separate two-sample t-test will allow you to test each covariate individually. Here, each parameter will be uniquely defined. Best wishes Martin > On 17 Jul 2015, at 15:45, Alexandre Obert <[log in to unmask]> wrote: > > Dear Martin, > > Thank you for your reply, I saw some information about this but I don't understand how to apply this. > Is this an option somewhere in SPM ? > And what could be the difference between this orthogonalization and setting two one-sample : one with the covariate1 and the second with the covariate2 (belong the fact that it will not be possible to test for both covariates together) ? > > Regards, > > Alexandre > > Le 17/07/2015 17:36, Martin Dietz a écrit : >> Dear Alexandre, >> >> If your covariates are correlated, you could orthogonalise the second with respect to the first. You can then test each of the covariates or both if this makes sense conceptually. >> >> I hope this helps >> >> Martin >> >>> On 17 Jul 2015, at 13:18, SUBSCRIBE SPM Anonymous <[log in to unmask]> wrote: >>> >>> Dear all, >>> >>> I would like to perform a correlation analysis from a one-sample t-test that contains cond1>cond2 con*.img files. >>> My question concerns the covariates I have to enter : they are correlated, is there any way to control this in the one-sample t-test design ? >>> Following this, is it possible to see the activations associated to the covariate 1, covariate 2 and both of the covariates in the one-sample design ? >>> >>> Regards, >>> >>> Alexandre Obert >