Hi Robin,
yep, the file melodic_IC contains unthresholded Z-maps - with the
--Ostats option you will also get thresholded maps and probability maps
ta
christian
On Thursday, May 29, 2003, at 21:05 Europe/London, Robin Goldman wrote:
> so if I want to average the unthresholded z-maps, is this what is
> contained in melodic_IC.img? Or do I need to run melodic with the
> "output stats folder" option and the threshold set to zero?
>
> On Saturday, May 24, 2003, at 10:22 AM, Christian F. Beckmann wrote:
>
>> Hi
>>
>> melodic Z-maps are signed, probability-maps are not since they show
>> different things: Z-maps show the number of std. deviations a given
>> regression coefficient is away from the voxel-specific noise - the
>> sign
>> just indicates if it's a positive or a negative regression
>> coefficient.
>> The probability map shows the significance of the finding based on an
>> alternative hypothesis approach.
>>
>> thanks
>> Christian
>>
>> On Friday, May 23, 2003, at 17:49 Europe/London, Vince Calhoun wrote:
>>
>>> sorry...the last message wasn't finished....
>>>
>>> regarding christian's suggestion...i believe the probability maps are
>>> signed (at least the z maps are...christian perhaps you could comment
>>> here as i don't want to put words into your mouth?), if so this
>>> wouldn't
>>> be a problem. if they are not signed it seems a straightforward
>>> extension to include the sign.
>>>
>>> cheers,
>>>
>>> vince
>>>
>>> Robin Goldman wrote:
>>>
>>>> Date: Fri, 23 May 2003 17:39:19 +0100
>>>> From: Robin Goldman <[log in to unmask]>
>>>> Subject: [FSL] more on ICA group maps
>>>> To: [log in to unmask]
>>>>
>>>> Hi all,
>>>>
>>>> So I've been mulling over ways to do an ICA group analysis, reading
>>>> through the past
>>>> postings (thanks!) and chatting about it all with Mark (Cohen). I'm
>>>> thinking there are three
>>>> ways I could go about this using MELODIC.
>>>>
>>>> First, a bit about the data set. It's more resting state data
>>>> (remember this? EEG/fMRI of resting
>>>> state to look at regions associated with alpha rhythm). I've got
>>>> multiple subjects, some
>>>> subjects have only a single scan and some have multiple scans.
>>>>
>>>> So I figure I could do this a number of ways, all of which have
>>>> their
>>>> statistical issues.
>>>>
>>>> 1. I like the idea of concatenating the data into one long time
>>>> series
>>>> to find a set of group
>>>> components. The probelm here is that some subjects have only one
>>>> scan,
>>>> while others have
>>>> up to 5 and I don't want to bias the group map towads these people.
>>>>
>>>> 2. Run ICA on each scan individually and average the zstat maps for
>>>> a
>>>> particular component.
>>>> But is this really statistically sound??? And again, the same
>>>> multiple
>>>> measure issue (I suppose
>>>> I could get a per-subject average z and then average these? Now the
>>>> stats get worse...).
>>>> Christian -- you suggest using the probability maps, but I want to
>>>> retain sign information. I
>>>> suppose I could multiply the probability maps by the sign from the
>>>> zmaps... but the stats on
>>>> averaging the probability maps seem odd to me too...
>>>>
>>>> 3. Merge the two methods -- concat for each subject, thereby getting
>>>> one IC-zstat map for
>>>> each subject, then average these.
>>>>
>>>> What do you think?
>>>>
>>>> Thanks.
>>>> Robin
>>>
>>> =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
>>> Vince D. Calhoun, Ph.D.
>>> Director, Medical Image Analysis Laboratory
>>> Olin Neuropsychiatry Research Center, Institute of Living
>>>
>>> Assistant Clinical Professor
>>> Department of Psychiatry
>>> Yale University
>>>
>>> Assistant Professor
>>> Department of Psychiatry
>>> Johns Hopkins University
>>>
>>> MAILING ADDRESS:
>>> Institute of Living
>>> 200 Retreat Avenue
>>> Whitehall Building #203
>>> Hartford, CT 06106
>>> Tel: (860) 545-7768, Fax: (860) 545-7797
>>> Email: [log in to unmask]
>>> =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
>>> "An author [or researcher] should never conceive himself as bringing
>>> into
>>> existence beauty or wisdom which did not exist before, but simply and
>>> solely as trying to embody in terms of his own art some reflection of
>>> eternal Beauty and Wisdom...And always, of every idea and of every
>>> method the Christian will ask not, 'Is it mine?' but 'Is it good?'"
>>> ----CS Lewis----
>>
>>
>
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Robin Goldman, Ph.D.
>
> Hatch Center for MR Research
> Columbia University
> 710 W. 168th Street
> NIB-1, Mailbox 48
> New York, NY 10032
>
> (212) 342-0867 phone
> (212) 342-5773 fax
> [log in to unmask]
|