Dear Dr Winkler, Dear Paul,
Thank you so much for your kind input. Indeed, very helpful! I reviewed many articles about "contrast of parameter estimates" to improve my shallow knowledge in this area. I highly appreciate your responses.
Thanks
John
Hi John,
The cope is a "contrast of parameter estimates", that is, a contrast of betas, or in yet other words, a weighted sum (such as a difference) between betas. That's all.
I see no reason to want to have them all positive... in fact, they can be negative.
All the best,
Anderson
On 18 April 2018 at 17:49, paul mccarthy <[log in to unmask]> wrote:
Hi John,
The Oxford Neuroimaging Primers book series has a set of free online appendices, one of which is on the GLM:
http://www.neuroimagingprimers.org/online-appendices/
Cheers,
Paul
On 18 April 2018 at 19:05, John anderson <[log in to unmask]> wrote:
Dear Dr Anderson– as always I highly appreciate your direct, friendly and fully detailed responses!! Thank you so much!
Kindly, regarding my question #3 below:
Honestly, I don't fully understand how the statistical tools (e.g. randomise or FSL/FEAT) handle these cope/varcope images. For instance, let's say we have positive signal in a specific voxel in image A and the signal is negative in the same voxel in image B and so on for images from A to Z. How the differences in the signal in this specific voxel will be generated between all the images? Kindly is there any reference that I can study to understand the copes images and the stats behind it?
I aim to convert the "copes" to images has positive values. Or vice versa generate "cope" images from another type of images– I am working on a multimodal neuroimaging analysis (fMRI and PET). I would like to use the same statistical approaches for both types of images. I analyzed fMRI in FSL/FEAT. The final output is copes. How can I generate copes from PET images? Or how can I convert the fMRI copes to images have positive values similar to the PET images. I would like to employee similar statistical analyses (fMRI and PET) using the same type of images (i.e. images have positive values. Or images have positive/negative values).
Any clarifications/advices are highly appreciated!
John
Hi John,
Please see below:
On 12 April 2018 at 07:08, John anderson <[log in to unmask]> wrote:
Dear FSL experts,
I would like to use "Randomise" to study the difference between two groups. I have COPE images as a final step of processing. these images are in the standard space. I merged all these COPEs using "fslmerge" then I built design matrix and design contrasts. Then I fed the COPEs-merged file to randomise as follows:
randomise -i COPEs -o output.nii -m mask.nii -d design.mat -t design.con -n 10000 -T -v 5
Kindly, I would like to inquire about:
1 Given that the COPE images contain positive and negative values. Is is correct to feed these images to randomise?
Yes.
2. Is it correct that FEAT FSL choose parametric approach to study the COPE images instead of non parametric? If yes, I highly appreciate if you clarify why?
FEAT uses parametric methods. This was the standard method when running a resampling method would take forever.
3. Is these any way in FSL to convert these COPE images to another type of images contain only positive values?
Sure we can add some constant but why would it be necessary?
All the best,
Anderson
Looking forward to learn from your experience!
I appreciate any advice!
Cheers: John
|