On Thu, 27 Jan 2005 13:03:52 +0000, John Ashburner <[log in to unmask]>
wrote:
>It's a stats question, so I left it for the guys who know about stats, whereas
>they saw it as a morphometry question, so left it for me. As a rule, if you
>want an authoritative answer about stats, it is better not to mention that
>you are doing VBM.
thnk you very much John, sorry but I'm not very familiar submitting to the list
>I'm not 100% sure what this design matrix looks like, but I suspect that the
>first two columns indicate which subjects belong to the two groups. In taht
>case, contrasts of [1 -1 0 0 0] and [-1 1 0 0 0] can be used to answer this
>question..
the design is as you suspect ;)
>Do you mean look for an interraction between group membership and the
>contrasts? If so, then you'll need a slightly different design matrix, which
>models the covariate seperately for each group. I'm not sure what options in
>the stats UI will give you this.
again, many thanks
Regards
Riccardo
|