ICA component topography after clustering - Document for release
arnodelorme opened this issue · comments
The issue is that replotting a component from the dataset or from a cluster returns different results. See scalp topography.
The issue was due to a modification performed in October 2023 for Octave compatibility, where the meshgrid function was used to precompute point coordinates. This is because Octave griddata function is not as powerful as the MATLAB one. However, the coordinates were inverted resulting in the bug above. The image below shows the same plot as above after the bug was fixed in fb2b77e
Important notes:
- This only affects replotting saved components
- If you used scalp topographies for clustering components, it is unaffected by this bug (this is because all scalp topographies are affected in the same way).
- Localization of components is unaffected (it is a different process)
- Plotting scalp topographies of ERPs and plotting components is not affected (this issue only arises because of how components scalp topographies are saved).
- If your scalp topographies were computed before 2024, then the issue will not arise.
- The only difference is that the cluster plot is wrong