Hi AFNI experts,
I wanted to compare the results of using different contrasts for the psychological variable and also ensure that my understanding of the PPI term interpretation is correct.
The current recommendation is to use contrasts of 0 and 1 for the psychological variable, which allows the interpretation of the PPI term to be a change in the relationship between the seed region and target region for a given condition compared to baseline.
If one was interested in using effects (deviant) coding of 1 for the condition of interest and -1 for the control condition, would the interpretation of the PPI term be the change in the relationship between the seed region and the target region for a given condition compared to the control condition?
Also, if using effects coding, If we had 4 conditions of the n-back (i.e., 0-back, 2-back, 3-back, and 4-back), would a fourth psychological variable be needed that codes all of the conditions as 1 as recommended by FSL [fsl.fmrib.ox.ac.uk/fsl/fslwiki/PPIFAQ]?
Lastly, we are ultimately interested in any linear effects of the n-back (i.e., does the relationship between the seed region and target region change as the n-back load increases?). I was thinking that this could be achieved using polynomial contrast coding, where the coding scheme of the four psychological variables would be:
PSY term 0back 2back 3back 4back Average 0.25000 0.25000 0.25000 0.25000 Linear -0.63246 0.00000 0.31623 0.63246 Quadratic 0.53452 -0.53452 -0.26726 0.53452 Cubic -0.31623 0.00000 -0.63246 0.31623
If so, would the subsequent PPI interpretation be as follows?
Average: The relationship between the seed region and target region on average across the contrasts
Linear: The change in the relationship between the seed region and target region as n-back load increases linearly