Dear Edward,
I feel like I've missed an email here or something, but I'll try and answer the best that I can…
> FSL -
>
> Each subject has 12 EVs. For two subjects, one of their EVs has all 0s in the third column. Thus the main effect maps are blank and I've removed these subjects from group level analyses. The third column values are parametric modulators with expected value or prediction error rather than indication that a certain event occurred.
>
> 1) Is there an appropriate way to address this so that I do not have to exclude these subjects? (i.e. add 1 to every single value for every regressor in the third column; alternatively, put .001 in just one element in the 0 regressor)
Do you have contrasts of interest that depend on this EV? If so, then it is really not possible, under any statistical framework that I can think of, where you can use subjects that have no information in such an analysis (at least in any sense that is different from just excluding them). I would not recommend putting any non-zero values the EVs, as that would then extract inappropriate and non-comparable information from the data.
> 2) As mentioned I am using values derived from a standard TDRL model. As I understand it, in SPM the parametric modulator looks at variance above and beyond the main effect. Am I doing the same thing in FSL by putting the expected value and/or prediction error in the third column of the regressors?
I don't know enough about SPM to comment much about what is done there, but you can certainly have a parametric modulator with a GLM in FSL. However, without knowing more details about how you have set it up, I can't really comment on whether your model is correct or not. If you are still unsure, send us some more information.
All the best,
Mark
> 3) If not, is there a way to do this type of analysis in FSL?
>
> Cheers,
>
> Edward
|