Hi, All,
We're trying to reanalyze some old data which originally was analyzed before the modifications to 3dANOVA2. I understand that there have been some changes in the way the new version calculates t-values for contrasts specified by the -contr when the contrast does not sum to 0. We're noticing large changes in t value for p<0.005 (i.e. from 2.902 to 3.962) even for contrasts which _do_ sum to 0. As you might expect, using this new t value in a clustering analysis results in many fewer clusters than in our old analysis. How should we deal with this? Why might we have an inflated t value even when our contrasts sum to 0?
Thanks--any suggestions are welcome!
Emily Myers