Orthogonalising CSF, WM and 6 motion parameters

Submitted by Mamtis13 on Wed, 09/18/2013 - 00:50

 Dear experts,

Can you please tell me if, in DPARSFA, CSF, WM and the 6 motion parameters are orthogonalised against each other before these covariates are regressed out during pre-processing?

Thank you very much for your advice!
Regards, Mamtis

Submitted by ZHANG_RESTadmin on Wed, 09/18/2013 - 11:09

 dparsf is using matlab function of "regress" (multiple linear regression) to regress out nuisance signals. 
 since we have lots of regressors, the matlab uses Least square method to estimate beta (regression coefficients), before doing this, it DOES remove the dependence between the nuisance signals.


Submitted by lixiangcx on Fri, 01/03/2014 - 11:24

Dear rest-fmri team,  
1. If I have regressed out nuisance regressors (CSF, WM and the 6 motion parameters), do I have to add realign parameters as covariate in the statistic analysis?
2.  In two sample t-test, does the resolution of data and covariate must be 61*73*61? What should I do if they are not?
Thank you very much and best regards!

Submitted by ZHANG_RESTadmin on Wed, 01/08/2014 - 13:32

a1: usually don't need to do it. But now head motion has been raised more frequently as a problem because some researcheres have stated that it will cause spurious result of "group difference".  So some reviewers may ask you to use head motion parameter as covariates in group comparisons. 

a2: in two sample t test, which covariates you should use depends on situation.  If you use largest or mean or "overall estimates" head motion as regressors, you can just extract those values and input them into text covariates.   If you want to use another brain images as regressors, you have to put them in the covariates images.