Hi - the effect of the burnin is more important in bedpostX than it
was in bedpost, as it is not easy to get a good initialisation of the
model when there us more than one fibre population, so you are doing
a good thing by testing this. A good thing to look at is the
difference in how often you see crossing fibres (e.g. where is
mean_f2samples greater than 0.05). The amount of burnin you need
may well depend on your data, but more should be better! It is
obviously a trade off with how long you want to wait.
Cheers
T
On 27 Nov 2007, at 20:18, Ted Yanagihara wrote:
> After searching the archives I found that Mahinda Yogarajah asked a
> question similar to the one I have just come up with, but there
> doesn't seem to be any reply. So, I'll try again.
>
>
> I have been playing with the default settings for bedpostX to see
> if there are any major changes to the data. To do this quickly I
> have been running single voxel analyses on individual subjects.
> When I seed a voxel in the rostral ACC at a burnin of 1000 I get
> fibers that go from the frontal pole, cross the midline, and go to
> the contralateral frontal cortex. If I then use a burnin of 5000
> these fibers start going through the genu, but then stop and do not
> reach the contralateral frontal cortex.
>
> Does anyone have suggestions on the right way to decide the
> appropriate burnin? I only vaguely understand what the purpose of
> the burnin is, so it's hard for me to make any educated guesses.
> Also, increases in processing time are not much of an issue, at
> least for the time being.
>
> Thank you!
>
> ted
|