Print

Print
Hi John
[log in to unmask]" type="cite">
Dear Jonathan, Cyril
thanks for your answers.
Yes actually i meant parametric modulators (not regressors).
My main target is to find correlates for PB. However I also need to make sure
that the correlates are specific to PB and not to PA and potentially also not to PE.
Is it an acceptable way if I start with PB and then put PA and  PE?
Or shall I put PB at the end, with the risk that PA might absorb all variance?
What is the best way to do it?
We recently did something similar (http://cercor.oxfordjournals.org/content/early/2012/04/05/cercor.bhs090.abstract)

I you want to be absolutely certain that your effect is related to PB (and uniquely to PB) then it has to go last, i.e. anything else you thought of is explained away, yet you found something! putting it first will not allow you to conclude this as the shared variance related to PA and PE will be attributed to PB (doesn't mean the effect is not there, but you can't be certain that this is uniquely due to PB)

cyril