 |
 |
On 7/23/07, Christian Beckmann <[log in to unmask]> wrote: > > Hi > > I'll give it a try. Many thanks! I think I now (almost) understand. Apart from one point: > Therefore, any signal change detected under condition A is equal > to a+b (that's the total 'height' modelled). > Ok, so this is the logic step I find hard to understand. I can see that EV1 (a) and EV2 (b) are not orthogonal in the design matrix. Hence that EV1 can't be just (A-B). And that you therefore need to refactor the EVs to recover the 'real' A-B contrast. I guess where I am unclear is how one knows to equate condition A (and exactly A not 2A or 2A-B-C etc) with the total height (a+b) of the model. It looks like one needs some extra knowledge about how the higher level design matrix is processed? Is it simply that condition A is set as +1 in the EVs (but then do not the -1 values for conditions B and C contribute to the scaled height of the model)? Perhaps it is that I do not understand well enough how the (higher order) design matrices are processed in FSL. If there is a good reference source for this, please point me in the right direction! kind regards Sophie