Hi,
Yes - I think you're right the manual doesn't mention these. These
are the temporal and subject modes in a tensor decomposition (which
you didn't do). If you choose the temporal concatenation approach
(which you did) they are slightly harder to interpret - basically the
long timecourse from melodic_mix (ie the ICA timecourse for a given
component) is reorganised into a timeXsubjects matrix and PCA is run,
and that's what gives the Tmodes and subject weightings - or something
like that....Christian knows the details ;-)
Cheers.
On 20 Apr 2009, at 09:34, M Mather wrote:
> Perfect - that is just the information I was looking for. Thanks so
> much!
>
> Out of curiosity, what do the columns in the melodic_Tmodes
> correspond with?
> I apologize if I am asking something that is detailed in a manual
> somewhere
> - I have been searching for information on what the output files
> correspond
> with but have not been able to find it.
>
> On Mon, 20 Apr 2009 09:15:17 -1000, Steve Smith
> <[log in to unmask]> wrote:
>
>> Hi - you want the file melodic_mix which contains a column for each
>> ICA component, and each column is the concatenation across time of
>> all
>> subjects' temporal chunks of that component's full timecourse - so
>> split that up vertically and then you can compare that data
>> separately
>> for each subject against their own model. I guess they are demeaned
>> so
>> the model will also need to be.
>>
>> Cheers.
>>
>>
>>
>> On 20 Apr 2009, at 04:55, SUBSCRIBE FSL Anonymous wrote:
>>
>>> Sorry, I meant to ask what the output file melodic_Tmodes saves. It
>>> has as
>>> many rows as there are volumes in my analysis, so it seems to be the
>>> time
>>> series, but although the number of columns increases as the number
>>> of
>>> sessions entered in the analysis increases, it is not a clear
>>> correspondence
>>> between the number of columns and the number of sessions (e.g., 234
>>> columns
>>> for an analysis with 72 sessions that yielded 15 components).
>>>
>>> thanks,
>>> Mara
>>>
>>> On Mon, 20 Apr 2009 15:34:25 +0100, M Mather <[log in to unmask]>
>>> wrote:
>>>
>>>> I have run a group ICA using the concatenation mode and have what
>>>> looks like
>>>> a meaningful set of components. However, we counterbalanced task
>>>> block order
>>>> across subjects and so cannot use the built-in post-stats options
>>>> to
>>>> correlate the components with what the subjects saw. Is there some
>>>> other
>>>> way we could compute a correlation between the time course of a
>>>> particular
>>>> component for a subject and the conditions that subject saw over
>>>> that time
>>>> course? It seems that we might be able to use the output files
>>>> from the
>>>> analysis if there is a text file in which each subjects' time
>>>> course for
>>>> each component was saved.
>>>>
>>>> Related to this, what is the information saved in melodic_Smodes?
>>>> I can't
>>>> figure it out since the number of columns is not divisible by
>>>> either the
>>>> number of components or the number of sessions.
>>>>
>>>> Many thanks for your help,
>>>> Mara
>>>
>>
>>
>> ---------------------------------------------------------------------------
>> Stephen M. Smith, Professor of Biomedical Engineering
>> Associate Director, Oxford University FMRIB Centre
>>
>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>> +44 (0) 1865 222726 (fax 222717)
>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>> ---------------------------------------------------------------------------
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|