Dear Yu,
Gareth is probably the best person to answer this, but note that
priors do not impose hard constraints and the solution can be far from
the prior if the likelihood favours that. It's just less likely (and
that's what you seem to be getting).
Best,
Vladimir
On Tue, May 31, 2016 at 10:43 AM, YU LI <[log in to unmask]> wrote:
> Dear SPM experts.
>
> I was trying to use VB-ECD in SPM12 to get the best source locations. The
> prior location variance (mm2) is [36 36 36], the prior moment variance
> (nAm2) is [100 100 100]. Predictably, the post locations should be within
> the sphere with 6 mm radius. But two of them are located out of the sphere.
> Further trials confirmed that it is unrelated to the number of iteration.
>
> Do you know what happened? Any comments or suggestions will be much
> appreciated.
>
> Thank you,
> Yu
>
>
> --
> Yu Li
>
> PhD candidate,
> Department of Cognitive Science
> ARC Centre of Excellence in Cognition and its Disorders (CCD)
> Macquarie University, Australia
|