Hi Wei-Ta,
I think I need to hear more about your procedures because it's not clear what you're doing. If you're using MELODIC's ICA combined with dual regression (as you've indicated), then it's unclear how FEAT is coming into this discussion. But, if you're conducting a seed-based analysis through FEAT (which would not involve ICA or dual regression), the group-level time series plot would indicate the beta weights for the peak voxel across subjects.
David
On Feb 1, 2013, at 10:17 AM, 陳韋達 <[log in to unmask]> wrote:
> Dear David:
>
> Thanks for your answers and Question 3 is now solved. As for Questions
> 1 and 2, I think there is some misunderstanding and I apologize for
> that. These two questions refer to the FEAT output rather than the
> MELODIC output. I need you to confirm if the data points of the
> time-series plot in the FEAT output are beta weights, z or others?
>
> Thanks again.
>
> Best,
>
> Wei-Ta
>
>
> 2013/2/1 David V. Smith <[log in to unmask]>:
>> Hi Wei-Ta,
>>
>> I'm not sure how you have output from FEAT if you're using MELODIC ICA and
>> dual regression. In any case, the report.html probably won't be of much use
>> to you since you're doing dual regression, but here's some feedback on your
>> questions:
>> 1) They're not beta weights. Assuming you're using concat, the displayed
>> time course is an approximation of all inputted datasets (see
>> http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/MELODIC for a better description).
>> 2) Sorry, I'm not sure about this one, but perhaps it is explained in the
>> documentation or prior posts on the listserv.
>> 3) Those numbers reflect the center of gravity of each cluster in MNI space
>> (usually only done with FEAT output, so are you sure you're using ICA and
>> dual regression?). The max COPE and max z-stat do not have to be the same
>> voxel -- e.g., a high COPE value with a high variance will have a low
>> z-stat.
>>
>> Hope that helps.
>>
>> Cheers,
>> David
>>
>>
>> On Feb 1, 2013, at 4:54 AM, 陳韋達 wrote:
>>
>> Dear David:
>>
>> Yes. I am using dual regression. Sorry I did not mention that previously.
>>
>> Do you have any clue or references for my questions? Thanks in advance.
>>
>> Best,
>>
>> Wei-Ta
>>
>>
>>
>> 2013/2/1 David V. Smith <[log in to unmask]>:
>>
>> If you're interested in group differences in networks estimated with ICA,
>>
>> then you should use dual regression:
>>
>> http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/DualRegression
>>
>>
>> David
>>
>>
>> On Feb 1, 2013, at 12:37 AM, 陳韋達 wrote:
>>
>>
>> Hi FSL experts:
>>
>>
>> I am analyzing resting-state functional connectivity data using ICA
>>
>> and cluster-based thresholding. Two groups are studied (patients
>>
>> [n=18] vs. controls [n=23]). When I check the higher level FEAT
>>
>> results (report.html), there are time-series plots (of the peak Z
>>
>> voxel) for each contrast (control group, patient group,
>>
>> control-patient, patient-control, 4 contrast in total for my study).
>>
>>
>> My questions are:
>>
>> 1. What are the y values of the Time-series plots? Are they beta
>>
>> weights or z values? As I know, the x value (0-50) is the numbers of
>>
>> the subject.
>>
>> 2. When I click the plots for each contrast, two plots of full-model
>>
>> fit will pop up. Then I click the "raw data", a four-colume text file
>>
>> will pop up. In this text file, I think the values of the most left
>>
>> colume are the y values of the time-series plot (as question 1).
>>
>> Columes 2 and 3 seems to be contrast-specific values. But I have no
>>
>> idea what are the values in the most right columes.
>>
>> 3. When I click on a cluster of activation on the color rendered FEAT
>>
>> stats, a webpage of "cluster List" table will appear. I would like to
>>
>> ask what are the values of Z COG X (mm), Z COG Y and Z COG Z? Another
>>
>> question is: why the Z MAX coordinate and COPE MAX coordinate are
>>
>> different?
>>
>>
>> Thanks!!
>>
>>
>> Best,
>>
>>
>> Wei-Ta Chen, MD, PhD
>>
>> Taipei Veterans General Hospital
>>
>>
>>
>> --------------------------------------------
>>
>> David V. Smith, Ph.D.
>>
>> Postdoctoral Fellow, Delgado Lab
>>
>> Department of Psychology
>>
>> Rutgers University
>>
>> Newark, NJ 07102
>>
>> --------------------------------------------
>>
>>
>>
>> --------------------------------------------
>> David V. Smith, Ph.D.
>> Postdoctoral Fellow, Delgado Lab
>> Department of Psychology
>> Rutgers University
>> Newark, NJ 07102
>> --------------------------------------------
>>
----------------------------------------------
David V. Smith, Ph.D.
Postdoctoral Fellow, Delgado Lab
Department of Psychology
Rutgers University
Newark, NJ 07102
----------------------------------------------
|