Print

Print
Dear All,

I'd agree with Christian, that if your only reason for changing programs is thresholding, then its a bit strange and might raise questions with reviewers.

However, if your reason for switching is to use a model that is available in SPM, that is not available in FSL, then I think the change would be acceptable. For example, if you had repeated-measures and wanted to correct for violations of sphericity or the covariance between the measures, then you'd want to use SPM as SPM can attempt to correct for the violations and for the covariances. Likewise, if you prefer SPM's first-level models; but want to use randomise in FSL, then this is also acceptable. I don't think anyone would criticize you for using the statistical package that has the tools to build the model you want to use.

If you want to use thresholding as done in SPM, I have a beta version of a toolbox called peak_nii.m. 
It is available here: http://www.martinos.org/~mclaren/ftp/Utilities_DGM/peak_nii/

I plan on moving it over to NITRC soon: http://www.nitrc.org/projects/peak_nii

You will also need the attached file. In addition, I've attached a rough version of the manual.


Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
Postdoctoral Research Fellow, GRECC, Bedford VA
Website: http://www.martinos.org/~mclaren
Office: (773) 406-2464
=====================
This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED
HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is
intended only for the use of the individual or entity named above. If the
reader of the e-mail is not the intended recipient or the employee or agent
responsible for delivering it to the intended recipient, you are hereby
notified that you are in possession of confidential and privileged
information. Any unauthorized use, disclosure, copying or the taking of any
action in reliance on the contents of this information is strictly
prohibited and may be unlawful. If you have received this e-mail
unintentionally, please immediately notify the sender via telephone at (773)
406-2464 or email.


On Sat, Jan 11, 2014 at 5:25 AM, Christian F. Beckmann <[log in to unmask]> wrote:
Dear George,

In this specific case it seems technically possible to feed part of the output from one method as input into another. 
Overall I’d like to advise against using such plug-and-play approach and to check your specific reasons very carefully. 
Often the intricacies of the analysis are such that while it might be technically possible, it has associated methodological issues. In your case you seem to be in favour of a particular way of thresholding and associated way of reporting. In order to achieve this you’re now willing to change not just thresholding and reporting, but also higher-level _estimation_ of effect sizes and associated variances. If it’s only about thresholding and reporting then I’m sure there is a possibility to feed the output from higher-level FEAt into matlab and call the relevant SPM thresholding routines...
 If I were a reviewer of any associated work I’d find your approach ill motivated. In many cases such ‘fiddling’ with the analysis chains points to findings not being robust and heavily dependent on small tweaks. This in turn suggests that you’re not powered adequately. Also, when you say ‘sensitive’ below you actually mean ‘lower p’/‘bigger blobs’  which does not speak to sensitivity per se. When you do things differently then yes, results differ - but it’s not the case that just values change; you are changing from FWE corrected results to FDR with different associated interpretations (i.e. the p value does no longer mean the same thing). 

hth
Christian


On 11 Jan 2014, at 09:02, G Ch <[log in to unmask]> wrote:

Thank you Jesper. That was my hunch too.

Sara, the results I'm getting 'make sense'. They are very similar to randomise output, but the stat values are a bit more sensitive. Interestingly both seem more sensitive than FLAME but I imagine that also depends on the data you have.

Best,

George

On Jan 11, 2014 6:22 AM, "Sara Yahoo" <[log in to unmask]> wrote:
Dear George,
Your problem was also a concern for me...would you please let us know about your final result for using the cope in SPM? 
Thank you...




Sent from my ipad.

On Jan 11, 2014, at 1:58 AM, Jesper Andersson <[log in to unmask]> wrote:

Dear George,

I did my first level Glm analysis in Feat but I would like to do my second level analysis in spm. So i can output cluster.fdr p values. Can I simply use the cope contrasts or do I have to diivide them by varcope? Would it be a problem to do the first and second level analyses in different softwares?

I don’t think that feeding the copes from FSL into SPM would present any problems. SPM uses an OSL approach, so you should just feed the copes in.

You might also want to ask the same question on the SPM helpline as they are the experts when it comes to what you need to feed into a 2nd level analysis in SPM.

Good luck Jesper


Thanks in advance.

Regards,

George