Hi Kathy,
I am not sure whether correlation is your problem, since grey columns
can happen if a scan belongs to more than one condition (if I remember
correctly).
However in the past I was able to get estimation for regressors with
correlation as high as 0.68 with parametric modulations. When reaching
r > 0.8 the estimation started to fail. It depends also on the number
of correlating factors, because the software needs to find a
non-correlating subspace for each regressor. The higher the number of
regressors to compete in the game, the more difficult is to find a
subspace for all of them.
Dorian
Il 24 aprile 2012 14:47, Kathy D. Gerlach <[log in to unmask]> ha scritto:
> Hi everyone,
>
> Does anybody happen to know what the level of collinearity between
> regressors needs to be for them to be characterized as 'not uniquely
> specified' (aka gray) in a first-level SPM 8 design matrix?
>
> Thanks so much!
> Kathy
|