Dear Buyean,
Note that the vectors are not related to the statistics, but simply
the summary
of the coordinate change between the two groups as defined by your
initial
EV (1st EV). Hence they are determined only by the grouping defined
in EV1 and the coordinates. They are independent of the statistics,
and with
the current setup you cannot distinguish between positive and negative
correlation, as it is testing the movement of coordinates in 3D and so
"positive"
and "negative" are hard to define in that case.
You will only see vectors where the statistics are significant. All
others are
suppressed.
As for the difference between outward and inward with using TICV only
- that
is largely because in that case TICV will be used to define the
groups, and so
you will have a very different definition of the groups (I imagine,
unless there
is a very strong correlation of TICV and the group).
Note that the direction of the vectors is explained in the
documentation, and
relates to your group EV, so hopefully this makes sense in your study.
All the best,
Mark
On 24 Nov 2009, at 23:48, Buyean Lee wrote:
> Hi Mark and Brian,
>
> I ran a vertex analysis to test group (1st EV) difference in the
> shape and volumes with age (2nd EV) and TICV (3rd EV; total
> intracranial volume) included as covariates.
>
> The Native_*_L_Caud3.vtk file, which represent correlation between
> caudate and TICV, showed very strong (blue) correlation between
> shape/volume and TICV, but the direction of the vectors (the small
> arrows) seem to be opposite what I expected; the arrows are pointing
> inwards. In contrast, the vectors (of Native_*3.vtk files) of all
> other structrues are outward.
>
> For your reference, the vectors are outward when I ran the
> correlation between vertices of caudate and TICV (only EV).
>
> I will appreciate any advice for this.
>
> Thank you,
>
> Buyean
|