I guess I'm missing a few details, as I'm not sure how this could be
set up in Glm.
Lets say I have 20 subjects and for each subject I extracted the mean
timeseries in two seeds, which are now in txt files. So I have 40
different timeseries files. I just want to get a correlation
coefficient for these two seeds. How does randomise or Glm work w/a
single time series? Ultimately I want to get a correlation coefficient
for each subject, which after transforming to a z-score, could be used
to predict a behavioral variable.
thanks
On Fri, Jun 24, 2011 at 12:14 PM, Christian F. Beckmann
<[log in to unmask]> wrote:
> Hi
>
> You can simply use Glm (or GLM_gui on a mac) to create the design and contrast matrix and then use randomise for test this design on your data.
> hth
> Christian
>
> On 24 Jun 2011, at 13:30, Eric Fine wrote:
>
>> Hi,
>>
>> I've seen a few post regarding seed-based resting-state analyses and I was just looking for a bit more detail.
>>
>> I've already extracted the timeseries for various seeds of interest, plus nuisance variables (CSF, white matter, global mean signal). I want to do two things: (1) correlate the timeseries between two seeds, after regressing out the nuisance variables plus motion parameters; and (2) create whole-brain correlation maps for one seed, again regressing out nuisance variables.
>>
>> I am assuming fsl_sbca would work for this, but I'm still unsure about the specific options I'd use.
>>
>> Thanks in advance for the assistance.
>>
>> Best,
>>
>> Eric
>
|