Hi Rodrigo,

Please see below (at the far end):

On 8 August 2017 at 16:56, Rodrigo Perea <[log in to unmask]> wrote:
cat design_Lcline.con
/ContrastName1  NC>AD
/ContrastName2  AD>NC
/NumWaves       3
/NumContrasts   2
/PPheights              1.521264e+00    1.521264e+00
/RequiredEffect         2.230   2.230

/Matrix
1.000000e+00 -1.000000e+00 0.000000e+00
-1.000000e+00 1.000000e+00 0.000000e+00


$cat design_Lcline.mat
/NumWaves       3
/NumPoints      38
/PPheights              1.000000e+00    1.000000e+00    7.592400e+00

/Matrix
1.000000e+00    0.000000e+00    1.471000e+00
1.000000e+00    0.000000e+00    1.559900e+00
1.000000e+00    0.000000e+00    7.019400e+00
1.000000e+00    0.000000e+00    1.094100e+00
1.000000e+00    0.000000e+00    1.175200e+00
1.000000e+00    0.000000e+00    2.241200e+00
1.000000e+00    0.000000e+00    6.344000e-01
1.000000e+00    0.000000e+00    1.143200e+00
1.000000e+00    0.000000e+00    2.900700e+00
1.000000e+00    0.000000e+00    7.859000e-01
1.000000e+00    0.000000e+00    1.299200e+00
1.000000e+00    0.000000e+00    1.278100e+00
1.000000e+00    0.000000e+00    1.627700e+00
0.000000e+00    1.000000e+00    1.176400e+00
1.000000e+00    0.000000e+00    1.368300e+00
1.000000e+00    0.000000e+00    2.451400e+00
0.000000e+00    1.000000e+00    2.760300e+00
0.000000e+00    1.000000e+00    2.908000e+00
0.000000e+00    1.000000e+00    2.772700e+00
1.000000e+00    0.000000e+00    1.241900e+00
0.000000e+00    1.000000e+00    1.460600e+00
1.000000e+00    0.000000e+00    1.294700e+00
1.000000e+00    0.000000e+00    2.256000e+00
0.000000e+00    1.000000e+00    1.266600e+00
0.000000e+00    1.000000e+00    2.916700e+00
0.000000e+00    1.000000e+00    2.237200e+00
0.000000e+00    1.000000e+00    2.069900e+00
0.000000e+00    1.000000e+00    3.534700e+00
0.000000e+00    1.000000e+00    1.128400e+00
0.000000e+00    1.000000e+00    2.365000e+00
0.000000e+00    1.000000e+00    2.417600e+00
0.000000e+00    1.000000e+00    2.092200e+00
0.000000e+00    1.000000e+00    2.240400e+00
0.000000e+00    1.000000e+00    3.375700e+00
0.000000e+00    1.000000e+00    7.592400e+00
0.000000e+00    1.000000e+00    1.288000e+00
0.000000e+00    1.000000e+00    2.317000e+00
1.000000e+00    0.000000e+00    1.684300e+00





###PREVIOUS EMAILS:


Hi Rodrigo,

Before giving an interpretation we need to see the design and the contrasts.

All the best,

Anderson


On 1 June 2017 at 20:12, Rodrigo Perea <[log in to unmask]> wrote:
Hi,
I am trying to apply randomise in a 4D sample using the following command:

randomise -i $FILE -o rand_${FILE//.nii.gz} -d design_Lcline.mat -t design_Lcline.con -n 5000 --T2 -D

 I performed these in different metrics (e.g. FA, RD, AxD, etc) and I have 40 nonzero voxels in the same XYZ locations but when I look at the output of *corrp_tstat1/2.nii.gz output the number of voxels decrease (but not in the *tstat1/2*.ni.gz output). So... my questions/concerns:

1). Could somebody give me a intuitive information on why/how is this happening? I'm about to read the TFCE paper but some discussion on this issue will be greatly appreciate it.

These additional zeroes are those that have 1-p = 0, that is, p = 1, meaning they are not significant at all after correction. The distribution of p-values in the corrected images is skewed, even if the null hypothesis is true, so we expect an excess one 1, which appear in the 1-p as zero, as if these voxels were "missing".
 

2) These voxels are along a specific tract and perpendicular (or along) voxels  are not included so I was wondering if the --T2 option (and/or some other options can be recommended).

If its TBSS data, the recommendation is --T2. That's it!

All the best,

Anderson

 

Thanks in advance,
Rodrigo