Hi Anderson 

thank you very much for clearing it up.

Yes, I was looking for z-scores, basically standardized regression coefficients - indeed not to be confused with the z-statistics.

However, I am still unclear on the output of --out_z.

After some research, I found that the z-statistics is derived by the parameter coefficient divided by its standard error. 

However, the same formula is attributed sometimes to t-statistics.

You mentioned that the z-statistics are derived by transforming the t-statistics. Could you help me in understanding how this is done?

How are this z-statistic maps generated from the raw regression coefficient values? I was not able to find a formula.

Sorry for being so picky, but I am using this maps for some analyses and I would like to be completely clear on this issue.

Thanks for helping me so far, I am really looking forward to hearing from you.

Best

Lorenzo

On Sat, Jul 22, 2017 at 11:50 AM, Anderson M. Winkler <[log in to unmask]> wrote:

Hi Lorenzo,

The z produced with --out_z is a test statistic, but I think what you want is another kind of z, that is, a z that reflects the effect size, i.e., something akin to the z produced after a Fisher's r-to-z transformation.

For a seed-based analysis, this and various related functions are available in FSLnets. You'd replace the melodic_IC.nii.gz for a file with the masks that define the (non-overlapping) seeds.

Hope this helps!

All the best,

Anderson

On 20 July 2017 at 02:21, lorenzo pasquini <[log in to unmask]> wrote:

Dear Anderson

yes, thank you.

And sorry for the confusion, what I wanted to ask was whether --out_z is giving the z-transformed coefficient.

Just to confirm: if I wanted to compare maps of coefficients (beta-values) across groups derived from a seed to whole brain analysis on task-free fMRI data, and I wanted to use z-scores, the --out_z output is not the one to use... I should z-transform the derived maps on a second step.

Thanks

Lorenzo

On Tue, Jul 18, 2017 at 11:26 AM, Anderson M. Winkler <[log in to unmask]> wrote:

Hi Lorenzo,

Please see below:

On 17 July 2017 at 14:39, lorenzo pasquini <[log in to unmask]> wrote:

Dear fsl users

I have a question regarding the output of fsl_glm --out_z

It is said to deliver output file name for Z-stats.

No, it will deliver a map, saved to the file name you supply.

 

What is exactly meant here? 

Are those the z-transformed residuals?

No. These should be the t or F map converted to z.

 

If yes, shouldn't the z-scores be normally distributed around 0? 

No, because there may be true effects and/or false-positives.

Hope this helps!

All the best,

Anderson

 

This seems not to be the case with my data.

Any help would be more than appreciated.

Best

Lorenzo