 |
 |
Dear Anderson, Thanks a lot for complete respond, very helpful! Best regards Gayane Sent from my iPad On 6 Dec 2012, at 17:24, "Anderson M. Winkler" <[log in to unmask]> wrote: > Hi Gayane, > > No, not safe to put all subjects into a big analysis, although this doesn't mean that nothing can be done. If you are lucky and have balanced data across groups and scanners, i.e., the group indicator and the scanner indicator are ultimately orthogonal to each other, then yes, in this particular case it's possible to put all into a single analysis, making sure that scanner effect is included in the design matrix. In the nice scenario of balancedness, it will also ensure an unbiased study-specific template for the registration. > > However, if more subjects from a given group were scanned in one MRI system than in the other, then it's not possible to disentangle scanner effect from group effect, and you may want to choose the safest route, which in most cases (for multiple subjects) means be conservative and regress out the scanner effect first, and only then test the group effect on the residuals. Or carefully use orthogonalisation. > > Another possibility is to run two separate analysis, one for each scanner, then combine using meta-analytical methods. One that is very straightforward and you may want to have a look is Fisher's method: http://en.wikipedia.org/wiki/Fisher%27s_method. > > All the best, > > Anderson > > > 2012/12/6 Gayane Aghakhanyan <[log in to unmask]> >> Dear FSL experts >> >> Is it reliable to combine dataset from different magnetic field scanner such as 1.5T and 3T for VBM analysis? >> >> Thanks >> Gayane >