Hi Alejandro,
The simplest, and what would be my choice, is to show the t-statistic
map instead of z-score. The critical value to threshold it is shown at
the end of the randomise run. If you didn't write it down, no problem,
just threshold the corresponding corrp file at 0.95, and use it to mask
the t-stat file.
To really obtain a z-score, there is the command ttoz. The input images
for this are the contrast of parameter estimates (cbsfile) and the
variance of these estimates (varsfile). These images can be obtained by
fitting the model again with fsl_glm (see the output options of this
command). Alternatively, make a file full of ones and use it as the
varsfile, and the tstat itself as the cbsfile.
The critical threshold can be converted from the threshold reported by
randomise or, if it wasn't written down, the file can be masked as above.
All the best,
Anderson
Am 27.02.14 10:53, schrieb Alejandro CL:
> Dear experts,
> I am preparing some figures with the obtained results in VBM. I have seen the z-score is most used than p value. From my *_corrp_tstat1.nii.gz, how can I get the z-scores maps? How can I get obtain the z-score threshold?
> Thanks a lot,
> Alejandro CL
|