global signal and movements

Dear all,
 I would like to ask you a question on regressing out covariables using REST.

 
I am creating a covariables file, to feed into the regression function. in this file, I concatenate movement regressors and global signal and any other covariable I would like to remove. my question is whether this procedure is appropiate, since global signal values are much higher than those on movements. Do the covariates interact to each other during the process of regressing them out? should I regress out global signal in a step apart from movements?