Dear Stefan,
The ordinary least squares estimation in the GLM is based on all the regressors entered during model specification, resulting in one beta image per condition/regressor. Accordingly, different models based on different (combinations of) predictors will result in different beta estimates. E.g. adding the motion parameters will explain some remaining signal changes due to motion, but it's not just explaining some additional variance, it also affects the other estimates.
Defining contrast vectors occurs at a later stage, it's just combining the beta images, e.g. [-1, 1, 0, 0, 0, 0, 0, 0] really just means -1 * beta1 + beta2 + 0 * beta3 + 0 * beta4 + ...
Best
Helmut
|