Thanks for your response! The output I am getting from fslpspec is much
higher than I expected. Most IC components report the power in the
thousands from a GUI run. The images from fslpspec have values up to 7
digits, do I need to scale the output somehow?
Chris
On Jul 16 2008, Christian F. Beckmann wrote:
>Hi Christopher
>
>On 15 Jul 2008, at 21:01, Christopher Bell wrote:
>
>> I want to make sure I am calculating the frequency bins correctly
>> for an
>> output of fslpspec. The first volume would be (.5) * 1/total time?
>>
>> Our data has 177 volumes and a TR of 2 seconds. For instance the first
>> volume would be (.5) * 1/354s = 0.001412429 Hz and the last volume
>> would be
>> (.5) * 1/2s = 0.25 Hz. Is this correct?
>
>Yep!
>
>> I have also created a 176 volume
>> data; if the even number makes the calculation easier.
>
>
>
>> My second question is about the defl and symm options. I have not
>> tried
>> running anything with defl yet. What does it do? Is it only for
>> individual
>> ICA?
>
>The deflation approach extracts one component at a time (as described
>in Aapo Hyvaerinen's original work on FastICA - melodic is an
>implementation of this approach). In my experience this is far less
>robust than the symmetric approach where all components are being
>estimated simultaneously, that's why the default option is to use the
>symmetric estimation approach.
>
>> I was curious if MELODIC contained a way to concatenate in space
>> instead of time for a group analysis?
>
>nope, not in the current release version.
>cheers
>
>Christian
>
|