Dear SPM users,

 

I have a question concerning order of columns (conditions and regressors) in my 1^st level design matrix.

In our analysis we put additional regressors (valence and arousal of stimuli used in the event-related design) to regress out the variability that might be explained by these regressors, in order to have possibility to more “purely” interpret the conditions (representing basic emotions, which by their nature have different valence and arousal).

The Reviewer of submitted paper asked us to put those regressors (valence and arousal) prior to conditions in the design matrix. He/she claims that the order of columns matters, and asked us to manually change the order of columns in SPM.mat as during estimation the data are orthogonalized in a way in which the order of columns matters (and only putting the regressors first will allow us to interpret the results as impact of basic emotions).

 

Thus I have two questions:

1.      Is the reviewer right? Will putting the regressors prior to the conditions change the betas and/or statistics?

(I know that the regressors are orthogonalized within condition, one after another, when they are put as parametric modulators – and then the order is crucial. But I do not know if and how are the columns orthogonalized during “standard” (not PMod) GLM computations).

I would be grateful for some links / documentation and/or papers on it – so I can explore it and share with the Reviewer.

2.      Where, in the non-estimated SPM.mat, to change the order of columns, to do what the Reviewer asked for? Is it SPM.xX.X? Will SPM12 let me estimate such a model?

 

I’ll be grateful for you comments, suggestions and links.

Marek