Small differences

(Olof Lindberg) #1

Hi Martin and Beau,
I am running a shape analysis in which there are very small differences between groups - and at some points one group is larger than the other - in other places it is the opposite …
Using the old shape analysis mancova wizard … I find it difficult to interpret the results … what one really would like is some option (like in correlation) in which I can ask for one sided test … like is group 1 larger than group 0 and the opposite.

Is there any way of getting such results?


(Martin Styner) #5

Hi Olof
Sorry for the delayed answer.

  1. If I understand correctly the order of the groups is not consistent across the surface (as some regions have group A > group B and other regions otherwise). It would not make sense to use a one-sided test in such as setting. One sided tests should only be used if there is independent reasoning why group A > group B (or vice versa), e.g. if by design one group should be larger than the other.
  2. the shape analysis results (if you look at the mean difference) should tell whether and where group A > B or vice versa.