Hi,
> After running FIRST and getting a set of (very nicely) segmented sub-cortical structures, I am about to run vertex-wise correlations while accounting for 2 covariates. I generated my design matrices so that columns 1 and 2 are covariates, and column 3 is the variable of interest. I just have 3 quick questions:
>
> - Is it OK that the variable of interest is that of column 3 instead of 1? I found the online info a bit confusing with regard to this, and it seemed to me that the ordering might be relevant.
Yes, this is fine. The results of testing this variable will be labeled with a number 3, that is all. Each variable is tested separately (with the others treated as variables of no interest - or confounds) and each of the different results saved with a different number. So if you have three variables in your design matrix then you will get three outputs, one associated with a simple contrast for each variable (e.g. [1 0 0] for number 1, [0 1 0] for number 2 and [0 0 1] for number 3).
> - Is it true that, for each contrast automatically generated by FIRST, data for that EV will actually be separated into 2 groups (below and above the mean) and the difference between their means will be tested?
No, the statistics will be done without considering any groups. So if the values in the variable of interest are not indicator values then you will effectively test the regression of your variable. The groups are used to define the mean vectors for display purposes only. If you _want_ the difference between the group means then you just use a variable with indicator values (e.g. 1 and -1).
> - I prefer generating my DMs manually; will FIRST ignore the PPheights?
Yes, it will ignore this.
All the best,
Mark
|