help
help > RE: Comparing Fisher z score results from Conn to Z-stat reported from FSL?
May 27, 2020 11:05 PM | Alfonso Nieto-Castanon - Boston University
RE: Comparing Fisher z score results from Conn to Z-stat reported from FSL?
Hi Islay,
The exact answer depends on many things that are not clear from your description. Could you please clarify which Fisher z-score measures are you referring to (e.g. seed-to-voxel connectivity maps for a single subject as read for example from CONN's beta_*.nii files; average group-level Fisher z-score values from a one-sample t-test as read for example from a con_*.nii file or reported in the 'display effects' figures, etc.) and also provide a bit more details about what the FSL-reported values represent (e.g. they could be z-scores that have been reported as peak statistics, or perhaps as cluster-level statistics, or they could be actual z-score maps that have been shared summarizing a one-sample t-test across all subjects, etc.)?
Thanks
Alfonso
Originally posted by Islay Davies:
The exact answer depends on many things that are not clear from your description. Could you please clarify which Fisher z-score measures are you referring to (e.g. seed-to-voxel connectivity maps for a single subject as read for example from CONN's beta_*.nii files; average group-level Fisher z-score values from a one-sample t-test as read for example from a con_*.nii file or reported in the 'display effects' figures, etc.) and also provide a bit more details about what the FSL-reported values represent (e.g. they could be z-scores that have been reported as peak statistics, or perhaps as cluster-level statistics, or they could be actual z-score maps that have been shared summarizing a one-sample t-test across all subjects, etc.)?
Thanks
Alfonso
Originally posted by Islay Davies:
Hi,
I am using the Conn Toolbox to perform Seed-ROI analyses to replicate another study where the paper authors used FSL.
In their group level results they report Z-stats for these measures, which are on a different scale to the Fisher-z scores that Conn produces.
e.g. the Z-stat scores in their results range from -1 to +7, and the Fisher-z scores in my results for the same regions range from around -0.1 to +0.4. The pattern of results from both studies is comparable, but the scale is different, so it is difficult to really compare, i.e. how similar is a "Z-stat" of -1 (reported from FSL) to a Fisher-z score of -0.1 (reported from Conn)? And how comparable is a "Z-stat" of +7 to a Fisher-z score of +0.4?
I've never used FSL, but I read on the FSL forum page that FSL caters for different methods, so one way of processing the data might produce Z-stats and another method can be used to produce Fisher-z scores. Most studies seem to report Fisher-z scores for FC, so I am not sure why/how they would report Z-stats. I have contacted the authors, but am yet to receive a reply.
Does anyone know if it is possible to convert the Fisher-z scores to this Z-stat scale? Or would I need to run my whole analysis again in FSL in order to really compare. That would be a pain given I have ~400 subjects..., and I would need to learn a whole new tool.
Thank you for any advice you can offer.
Kind regards
Islay
I am using the Conn Toolbox to perform Seed-ROI analyses to replicate another study where the paper authors used FSL.
In their group level results they report Z-stats for these measures, which are on a different scale to the Fisher-z scores that Conn produces.
e.g. the Z-stat scores in their results range from -1 to +7, and the Fisher-z scores in my results for the same regions range from around -0.1 to +0.4. The pattern of results from both studies is comparable, but the scale is different, so it is difficult to really compare, i.e. how similar is a "Z-stat" of -1 (reported from FSL) to a Fisher-z score of -0.1 (reported from Conn)? And how comparable is a "Z-stat" of +7 to a Fisher-z score of +0.4?
I've never used FSL, but I read on the FSL forum page that FSL caters for different methods, so one way of processing the data might produce Z-stats and another method can be used to produce Fisher-z scores. Most studies seem to report Fisher-z scores for FC, so I am not sure why/how they would report Z-stats. I have contacted the authors, but am yet to receive a reply.
Does anyone know if it is possible to convert the Fisher-z scores to this Z-stat scale? Or would I need to run my whole analysis again in FSL in order to really compare. That would be a pain given I have ~400 subjects..., and I would need to learn a whole new tool.
Thank you for any advice you can offer.
Kind regards
Islay
Threaded View
| Title | Author | Date |
|---|---|---|
| Islay Davies | May 24, 2020 | |
| Alfonso Nieto-Castanon | May 27, 2020 | |