Print

Print
Dear Sir,

Thanks for the reply...

Sir can you please elaborate, as change in one of these DTI (FA,RD..) has influence on the other as reported in the literature. then it means that all these not independent? ..... please suggest me....

Also ..just  to know

In task based fMRI (task & rest type PPI) number of times we give a task and rest (let be 10times) to get the activation in specific regions.

Then  it means that we need to keep the p value (0.05/10 = 0.005), as we using all the 10 times fMRI task test... in the statistical analysis (to increase power...), In addition to correction for FDR/FWI for across voxels, for number of contrasts..

Please clarify

On Tue, May 17, 2011 at 8:22 PM, Stephen Smith <[log in to unmask]> wrote:
Hi

This is similar to the question in task FMRI of doing multiple comparison correction across different contrasts.

The strict answer is that yes, if the 4 tests were all independent then it would strictly be correct to add an extra factor of 4 to the correction. However, a) if they are not independent of each other then this is over-conservative and b) AFAIK no-one actually does do this…….

So your reviewer is both right and not so right…maybe Tom has more thoughts on this.

Cheers.




On 17 May 2011, at 03:33, Bhavani shankara wrote:

Dear FSL list

We have done TBSS analysis using all Fractional Anisotrophy (FA), Mean Diffusivity (MD), Axial Diffusivity (AD) and Radial Diffusivity (RD), we reported at multiple correction corrected results for all each of these using randomise at statistical significance level of P<0.05 (Corrected for multiple comparison FWE).

One of the reviewer commenting that we should have used (P<0.05/4= 0.0125 (for each of FA,MD, RD, DA) ) . I think he is suggesting the correction according to Bonfferoni correction.

My question is, is it sufficient to use p<0.05 FWE corrected for each of the DTI parameters or Should i use the significance level of p<0.0125 on the FWE corrected images as i am doing test for 4 times using FA,MD,RD and AD..

Please suggest me and clarify me what is the best statistical significance level to use...?

Thank you

--
Dr. BHAVANI SHANKARA BAGEPALLY
MBBS, (PhD in Clinical Neuroscience) NIMHANS,
Bangalore,
INDIA-560029
email:  [log in to unmask]
       


---------------------------------------------------------------------------
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]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------






--
Dr. BHAVANI SHANKARA BAGEPALLY
MBBS, (PhD in Clinical Neuroscience) NIMHANS,
Bangalore,
INDIA-560029
email:  [log in to unmask]