Hi
the model order (estimated number of components) does grow with the
number of time points - this makes sense because with more time points
you are able to sample different temporal dynamics, so yes, your
intuition is right. Limiting the number explicitly is an option in
case 40 components are too much to hande for whatever post-processing
you have in mind.
hth
Christian
On 21 Sep 2008, at 20:12, Christina Hugenschmidt wrote:
> Hi all,
>
> I have two data sets of resting state data I am analyzing
> separately. One has
> a TR of 3 and 100 volumes in a series and the other has a TR of 1.7
> and 200
> volumes. When I allows FSL to determine the number of components to
> output, there are usually between 10-20 components for the first
> data set
> (TR3/100 vols) and upwards of 40 for the second data set (TR1.7/200
> vols).
>
> I am assuming that the higher temporal resolution of the second data
> set
> accounts for this, but I was wondering if that is correct?
>
> Also, with that many components being output, should I be limiting
> the output
> to a smaller number?
>
> As always, thanks for any advice! The list is always most helpful.
>
> Regards,
> Christina
|