Dear Miss Engineering,
I'm afraid you cannot use that option unless your input data is
a 4D image (lots of 3D volumes) which represent the results of
statistics on appropriately permuted data. It is essentially a
very low-level randomise-like functionality (where randomise
is our tool for permutation-based non-parametric statistics).
If you only have a 3D image containing one z-value at each
voxel then this is not an appropriate option to use. If all you
have are z-values then you can only convert them to the
equivalent p-values. Whether the p-values are corrected or
uncorrected depends on how the z-values were created. If
they came from a FEAT analysis then they are uncorrected
and you can only use fslmaths to get uncorrected p-values.
All the best,
Mark
On 16 Sep 2011, at 17:13, Leslie Engineering wrote:
Just to clarify, to use my zstat file from group analysis to get corrected p value map, could I used a command like:
fslmaths zstat1.nii.gz -ztop -cpval pstat_image.nii.gz
On Fri, Sep 16, 2011 at 12:03 PM, Leslie Engineering
<[log in to unmask]> wrote:
Thanks!On Fri, Sep 16, 2011 at 11:57 AM, Mark Jenkinson
<[log in to unmask]> wrote:
Hi,
Actually, even this minimal documentation is unfortunately misleading/erroneous.
As the name suggests, you should pass in a z-value and get out a p-value.
The "<p>" in the usage should really be a "<z>".
$ ztop 1.65
0.049471
$ ztop 2.3
0.010724
$ ztop 3.1
0.000968
Jesper's suggestion, of using the fslmaths option, is more appropriate
for images containing z values. So choose whatever is most appropriate.
All the best,
Mark
On 16 Sep 2011, at 16:52, David Flitney wrote:
The tool (part of FSL) is documented as follows:
1 $ ztop
Usage: ztop <p> [-2]
-2 : use 2-tailed conversion (default is 1-tailed)
2 $ ztop 0.9
0.184060
3 $ ztop 0.9 -2
0.368120
On 16 Sep 2011, at 16:48, Leslie Engineering wrote:
I can't seem to find anything in the literature about ztop. Could you help direct me?
On Fri, Sep 16, 2011 at 11:24 AM, David Flitney <[log in to unmask]> wrote:
Don't know how to use fslmaths but single z-scores can be converted with ztop
--
Dave Flitney
IT Manager, FMRIB Centre
On 16 Sep 2011, at 16:20, Leslie Engineering <[log in to unmask]> wrote:
> How do I use fslmaths to do a statistical conversion from z value to p-value?
>
> THANKS!
