Print

Print
Hi Nebl,

Thanks for your help. I think I am starting to understand!

Yes, I think I have 3 regressors as you suggest! 

So now to set up the contrast in the SPM contrast manager. Do I understand correctly that I test the effect of each regressor separately? 

First, by putting in a T contrast [1 0 0], I test only the effect of ONLY my first regressor, the task activation (versus rest). (I attach the output figure, I guess the shaded regions show voxels where task is significantly different from rest?)

Then, by putting in a T contrast [0 1 0], I test for the effect of ONLY my second regressor, the behavioural parametric modulator.

If I were interested to test both first (task activation) and second (behaviour PM) together, how can I do this? Something like [1 1 0]?

Forward the corresponding con images (stored in the individual analysis folders) into a one-sample t-test for group statistics. 

Do you mean after having done the first-level analysis, to forward the con images into a 1 sample t-test for the second-level (group) analysis? How can I do this.. do I forward after starting the second-level analysis, or in the first?

Within the group models you would either set up an (undirected) F contrast [1], showing voxels associated with a significant effect, or go with T contrasts. As T contrasts within SPM are one-sided you would go with two contrasts within the group model, [1] showing positive task activations (or positive linear relationship wih the PM = increasing activations with increasing PM values) and [-1] for negative task activations (or negative linear relationships).

Here you also refer to the second-level group analysis right?

Thanks,
Joelle





On Fri, Jun 5, 2015 at 2:29 PM, H. Nebl <[log in to unmask]> wrote:
Dear Joelle,

No, the design matrices are created before the analysis. For "condition regressors" stimulus function matrices (stimulus on/off) are constructed based on onsets, durations, which are then convolved with e.g. the canonical HRF, resulting in the different predictors, which are then used in the GLM. In your case you probably have two regressors reflecting the task and the PM, one constant and possibly some additional regressors corresponding to the motion parameters (to explain some of the remaining variance due to head motion). From a statistical point of view there's no difference between "condition regressors", "PM regressors", or "regressors", it's just different ways how they are constructed.

For "task activation" you would set up a T contrast [1 0 ... ] on single-subject level with ... indicating zero-padding the rest of the columns. Forward the corresponding con images (stored in the individual analyis folders) into a one-sample t-test for group statistics. Same for linearly modulated activations based on the behavioral data, the PM should be represented by the 2nd row in your design matrix, thus you would go with [0 1 ... ]. Within the group models you would either set up an (undirected) F contrast [1], showing voxels associated with a significant effect, or go with T contrasts. As T contrasts within SPM are one-sided you would go with two contrasts within the group model, [1] showing positive task activations (or positive linear relationship wih the PM = increasing activations with increasing PM values) and [-1] for negative task activations (or negative linear relationships).

The SPM manual also has several chapters on how to set up single-subject and group models based on some example data sets, which can be retrieved from http://www.fil.ion.ucl.ac.uk/spm/data/.

Best

Helmut