help
help > RE: changes in v14
Jun 18, 2014 03:06 AM | Alfonso Nieto-Castanon - Boston University
RE: changes in v14
Hi Kathrin and Merina,
The main change that I believe could possibly have a noticeable effect on your results if you re-run the analyses on the new version is the modification to the provided Brodmann areas atlas. In version 14 we modified the default BA.img atlas included in the toolbox to improve its coregistration to MNI space by using the Lancaster transform (Lancaster et al. 2010). Depending on how your projects are set up (see Setup->ROIs) they might be pointing towards the older (v13) or newer (v14) BA.img atlas files, and this might have an effect on the observed results, although I would not expect any dramatic differences (this will affect seed-to-voxel and ROI-to-ROI analyses, but not voxel-to-voxel analyses).
Other potential differences that I would check, just in case, are whether the new default "detrending" and "despiking" options in the Preprocessing step (introduced around version 13n) are set similarly in the two analyses, whether the threshold values chosen in the second-level result explorer windows are equally set across the two analyses (for example we deprecated the use of FDR-corrected peak stats in favor of FWE-corrected peak stats, and we modified the default threshold values that you will get when launching the explorer), and whether you were using the same SPM version across the two analyses.
Other than the above, most modifications to the toolbox have been in adding new analysis options, but nothing that would change the main analysis results (other than very minor corrections or bug fixes which should have none or only relatively small impact on the analyses). In any way, please let me know if any of the above could explain your observed differences. If not, please let me know more details about how your new analyses were set up (whether you loaded the old conn_*.mat project using the new version of the toolbox or you recreated it from scratch; and whether the spatial preprocessing was re-run or used the same already preprocessed files) and send me your conn_*.mat files for the old and new analyses to further look into this.
Best
Alfonso
Originally posted by Merina Su:
The main change that I believe could possibly have a noticeable effect on your results if you re-run the analyses on the new version is the modification to the provided Brodmann areas atlas. In version 14 we modified the default BA.img atlas included in the toolbox to improve its coregistration to MNI space by using the Lancaster transform (Lancaster et al. 2010). Depending on how your projects are set up (see Setup->ROIs) they might be pointing towards the older (v13) or newer (v14) BA.img atlas files, and this might have an effect on the observed results, although I would not expect any dramatic differences (this will affect seed-to-voxel and ROI-to-ROI analyses, but not voxel-to-voxel analyses).
Other potential differences that I would check, just in case, are whether the new default "detrending" and "despiking" options in the Preprocessing step (introduced around version 13n) are set similarly in the two analyses, whether the threshold values chosen in the second-level result explorer windows are equally set across the two analyses (for example we deprecated the use of FDR-corrected peak stats in favor of FWE-corrected peak stats, and we modified the default threshold values that you will get when launching the explorer), and whether you were using the same SPM version across the two analyses.
Other than the above, most modifications to the toolbox have been in adding new analysis options, but nothing that would change the main analysis results (other than very minor corrections or bug fixes which should have none or only relatively small impact on the analyses). In any way, please let me know if any of the above could explain your observed differences. If not, please let me know more details about how your new analyses were set up (whether you loaded the old conn_*.mat project using the new version of the toolbox or you recreated it from scratch; and whether the spatial preprocessing was re-run or used the same already preprocessed files) and send me your conn_*.mat files for the old and new analyses to further look into this.
Best
Alfonso
Originally posted by Merina Su:
Hi Kathrin,
I have found the same thing, although not very dramatic differences - the significance values changed, which led to fewer clusters, but main results were still present...
I have found the same thing, although not very dramatic differences - the significance values changed, which led to fewer clusters, but main results were still present...
Threaded View
| Title | Author | Date |
|---|---|---|
| Kathrin Kollndorfer | Jun 17, 2014 | |
| Merina Su | Jun 17, 2014 | |
| Alfonso Nieto-Castanon | Jun 18, 2014 | |
| Merina Su | Jun 18, 2014 | |
| Alfonso Nieto-Castanon | Jun 18, 2014 | |
| Merina Su | Jun 19, 2014 | |
| Alfonso Nieto-Castanon | Jun 20, 2014 | |
| Merina Su | Jun 25, 2014 | |
| Eugenio Abela | Jun 25, 2014 | |
| Merina Su | Jun 25, 2014 | |
| Matthias Heil | Jun 25, 2014 | |
| Julia Landsiedel | Jun 25, 2014 | |