Print

Print
Hi Aleksandra,

This explains: TFCE considers the spatial signal as spread across multiple neighbouring voxels, something that SPSS cannot do, so neither the same F-test, nor the post hoc t-tests are guaranteed to be significant -- and often they are not, as TFCE is more powerful.

That said, even in randomise, the post hoc t-tests may still not be significant, as TFCE doesn't guarantee that the relationship between the F-test and its constituent (post hoc) t-tests are preserved.

All the best,

Anderson



On 5 May 2016 at 00:10, Aleksandra Klimova <[log in to unmask]> wrote:

Hi Anderson,

 

Yes, I did.

 

Many Thanks,

Aleks

 

From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Anderson M. Winkler
Sent: Wednesday, May 04, 2016 6:07 PM
To: [log in to unmask]
Subject: Re: [FSL] F test using randomize

 

Hi Aleksandra,

Did you use TFCE in randomise?

All the best,

Anderson

 

On 4 May 2016 at 08:11, Dr Aleksandra Klimova <[log in to unmask]> wrote:

Hi guys,

I have ran an F test using randomize which produced significant results. I then applied threshold to this F test and binarized it in order to create a mask. Finally, I extracted values in order to run post hoc tests in SPSS using the binarised F test mask. The Ftest is not significant in SPSS and neither are the posthoc contrasts. Any ideas why this could be the case?

Many Thanks,
Aleks