 |
 |
Hi Jinsu, > I have desgined paire t test > so I got reasonabl result > > However, I copied these files into another foleder > The same statistical analysis was then performed > > but, I found that result was not the same. > in addition, design matrix was also different although every setting was > the same. The first design matrix has the condition regressors before the subject ones (I believe SPM99/2 used to do this), while the second design matrix has the subject ones first (I think SPM5 does this). I can't think how this could happen without changing software version or similar, but it shouldn't be a major problem... The equivalent contrast for the second design of a contrast [a b] for the first, would be [zeros(1,N) a b] where N is the number of subjects. E.g. for 10 subjects, and a contrast of [-1 1] for the original design, you'd need [zeros(1,10) -1 1] for the second design. Hopefully the results are the same then. Note that SPM zero-pads to the right of contrasts which are too short (the [a b] of the first design becomes [a b zeros(1,N)] automatically) Best, Ged.