Hi Mirko,
> I have groups A and B, in which the mean values and
> large majority of individual VOIs when extracted and compared show A>B,
> but after proportional scaling the values are in revers order; A<B.
I think this just indicates that the means are more different than the
VOIs you are interested in, so when you proportionally scale by the
means, the VOI results are reversed.
Consider a simplified example, two element vectors, where the second
element is your "volume" of interest: a = [12 6], b = [5 4], so in
terms of the overall means (9,5.5) we have a>b, and in terms of the
VOI we have a>b (6,4).
But after proportional scaling (so that each vector has mean 1), we
have a = [1.33 0.67], b = [1.11 0.89], and so now in terms of the VOI b>a.
Best,
Ged.
|