Is an implementation of AR-IRLS planned to mne-nirs based on Fig.2 from the Huppert’s, 2016 paper (https://doi.org/10.1117/1.NPh.3.1.010401) comparing different models? It seems that AR(n) plus IRLS shows the best performance also in comparison to AR(1)-DCT.
Unlike in https://mne.tools/mne-nirs/dev/auto_examples/migration/plot_02_nirstoolbox.html, AR(n) was successfully implemented to run_glm/nilearn (https://github.com/mne-tools/mne-nirs/pull/277 - thanks for that:) ). However, I wonder whether IRLS provides any benefit in comparison to what is used right now?
- MNE version: 1.3.1
- mne-nirs: 0.5.0
- operating system: Windows 11
Best regards,
Brais