Dear Alfonso

Thank you so much for the wonderful CONN toolbox. I have found it easy to start using the toolbox with the help of the manual and this forum.
My question regards 2nd-level covariates. I have data from two group of subjects measured with two different head-coils, however the other coil was used more frequently i.e. the amount of measurements with different coils is not balanced. Both coils are used in both groups.
I would like to include the coil as a covariant in the analysis and remove its effect. I did a 2nd-level covariates GROUP1 [1111100000] and GROUP2 [0000011111] for group comparison.
Should I make a second level covariate (COIL) with subjects denoted as 1 for the other coil and 2 for the other(for example [1121112211]. Then, in the second-level analysis, when making a group comparison and controlling for the coil-effect, should I first select GROUP1, GROUP2 and COIL and the use a between-subject contrast of [1 -1 0].
Thank you in advance
Best regards
Jatta

Dear Jatta,

Yes, that is exactly correct. Selecting GROUP1, GROUP2, and COIL and entering the between-subject contrast [1 -1 0] will give you the desired between-group differences after correcting for potential connectivity differences between the two head-coils used.

If, in addition, you want to look at simple main effects (e.g. average connectivity in GROUP1 after discounting potential COIL differences between the two groups) you may want to "center" the covariate COIL (e.g.use  [-0.3 -0.3 0.7 -0.3 -0.3 -0.3 0.7 0.7 -0.3 -0.3] instead of [1 1 2 1 1 1 2 2 1 1]) so that the zero-level of this covariate represents the "average" effect across all your subjects (note: "centering" has no effect on the original between-group differences analysis).

Hope this helps
Alfonso

Originally posted by Jatta O:

Dear Alfonso

Thank you so much for the clarification and sorry for the late response. 

Now that I have found group differences I would like to extract individual data after regressing out the nuiscance effect of COIL and then plot these values in SPSS. The REX-toolbox, SPM8 and the calculator in CONN give nice figures of the results but I am struggling to extract the data plotted in the figures. 

I know now how to get the raw data out (that is, the "import connectivity values" method in seed-to-voxel and ROI-to-ROI windows'), but what I really would like to have are the individual fitted model values and the residuals. I mean, I would like to extract the subject-values after the effect of COIL is removed. Can this be done or is this data located somewhere?

Best regards
Jatta

Dear Jatta,

One possible way to get those fitted-values and residuals would be the following (assuming you already used 'import values' to get a second-level covariate containing the subject-specific connectivity values of interest; I am going to assume here that this covariate is named 'connectivity'):

1) first create a new second-level covariate in Setup.Covariates.2ndLevel (e.g. named 'connectivity_residual') copying the original second-level covariate of interest (this is just to avoid overwriting the original variable). To do this simply click on '+', enter the new covariate name 'connectivity_residual', and in the values field enter 'connectivity' (without the quotes) and hit enter.

2) then right-click (or command-click if on a Mac) on the new covariate 'connectivity_residual' and select the option 'orthogonalize selected covariate'. In the new prompt select the orthogonal factors 'AllSubjects' and 'COIL' and then click on 'Ok'. The new covariate 'connectivity_residuals' will now contain the desired residual values for each subject.

3) create a new second-level covariate (e.g. named 'connectivity_fitted') and in the values field enter 'connectivity - connectivity_residual' (without the quotes). This new variable will contain the desired fitted values for each subject.

Hope this helps
Alfonso 
Originally posted by Jatta O: