Dear Lisa,
With +/-0.5 you basically contrast the average of conditions 1, 2 and the average of conditions 3,4. With +/-1 you contrast the sums. The statistical results (p/T values) are identical, as it's just upscaling everything (M and SD) by factor 2. However, the con estimates in the con images are going to differ. This is only relevant if you want to extract values from the con images and/or if you plot some estimates within the SPM GUI. Of course, you can still adjust the scaling later on manually, but it's better to go with the correct weighting right from the beginning (as others might look at the models who are not aware of the initial weighting factors), and especially important, with the same weighting across subjects.
Best
Helmut
|