Hi,
I don't really understand your design.
If you are interested in whether there is a different slope (correlation)
for the relationship between geometry (vertex position) and the behavioural
measure in the two groups then you should have EV2 contain the
behavioural measures for group 1 and zeros for group 2, while EV3
should contain zeros for group 1 and the behavioural measures for
group 2.
The vectors (or meshes) will not be very meaningful for these latter
two EVs so I wouldn't bother trying to interpret them. Only the first
EV will give useful mesh differences and vectors (showing the
differences between the two group means).
I also hope that you have more than 6 subjects in your full analysis,
as it will be difficult to find statistically significant results will such
small numbers.
All the best,
Mark
On 16 Feb 2012, at 04:43, Chou Kun Hsien wrote:
> Dear Mark
>
> I have another following question, how to interpret the result of EV3 in vertex analysis ?
> How to interpret the two meshes and vector direction in the result of EV3 ?
> By the way, whether the design matrix which illustrate below is correct for interaction analysis in shape analysis?
>
> EV1 EV2 EV3
> -1 3 3
> -1 1 1
> -1 2 2
> 1 4 -4
> 1 5 -5
> 1 3 -3
>
> PS: EV1 is group variable (-1: normal ; 1: patient) ; EV2 (raw data of behavior mesure) ; EV3
>
> sorry to bother you !
>
> Best
>
> KH
>
|