The values are not from a regular GLM analysis, but represents classifier weights. Scale [-500,1500] means that the lowest weight is -500 and the highest weight is 1500. The highest weights typically appear in the same area of the brain for each subject and timepoint.
-Anders
________________________________________
Från: FSL - FMRIB's Software Library [[log in to unmask]] för Stephen Smith [[log in to unmask]]
Skickat: den 2 maj 2013 01:58
Till: [log in to unmask]
Ämne: Re: [FSL] Normalization of values
Hi - I'm afraid it's not clear to me what you mean by "scale" here?
Steve.
On 1 May 2013, at 15:47, Anders Eklund <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Dear experts,
I have volume data for a number of subjects and timepoints, and I would like to test if there is any group difference between different timepoints. The problem is that the data have rather different scales.
For example;
Subject 1, timepoint 1, the scale is [-500, 1200]
Subject 1, timepoint 1, the scale is [-600, 1500]
Subject 1, timepoint 1, the scale is [-400, 1800]
Subject 1, timepoint 1, the scale is [-700, 1100]
This makes it a bit hard to know if the change in value in a specific voxel is due to a real change or due to a scale change. What do you think is the best preprocessing option before I do a t-test with randomise?
1) Do nothing.
2) Normalize the values to be within the scale [0,1].
3) Normalize each volume to have a L1-norm of 1.
4) Normalize each volume to have a L2-norm of 1.
5) Other?
cheers,
Anders
---------------------------------------------------------------------------
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]<mailto:[log in to unmask]> http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|