Hi everybody,
My question may be simple but I have the following doubts: I fitted an ANCOVA, three groups of subjects and two covariates. I get an F contrast F(2, 38) that shows me a certain brain regions indicating group differences. I have different sample sizes for each group. However, when I apply t contrasts to test differences between each of the groups (for example 1 0 -1 to test if group A has a bigger response than group C), I get a t map with 38 degrees of freedom. For different group combinations I get always the same degree of freedom (38). I assume this is because I stay within the same model with the same number of parameters that had to be estimated.
But this is not the same as doing post-hoc t-tests (whole brain volumes) to test for differences between the different groups where the degrees of freedom change according to the group size.
What is the correct approach?
Thanks for your input and help,
Best,
Stephan
|