In running a glm analysis with multiple classes (A, B, C, D, E, F) and no variables, is it better to:
(1) have one fsgd with a contrast vector of (1 -1 0 0 0 0) to compare Group A thickness to Group B thickness, or (2) should I make an fsgd with just my Group A and B subjects and have my contrast vector (1 -1)?
We ran both options to compare Group A and B as an example and the results of each are extremely similar but there are some very minor differences in the significance maps. We're wondering about the two methods because we eventually want to compare every two-group pairing (A vs C, A vs D, E vs F, etc.), and was wondering whether one method is more accurate or preferred over the other when we try multiple group comparisons.
Also, could there be a randomizing effect that causes the slightly different maps for our example?
Thanks. David