[Mne_analysis] MVPA - analysis difficulty

[Ammara Nasim]

        External Email - Use Caution        

Dear MNE  and scikit-learn team,


I hope this email finds you well.


I am a Master's Neurocognitive student and I have recently started using MNE for my thesis. In my thesis, we are using an emotion recognition task and participants' EEG is being recorded. In the analysis, we will be decoding the emotions presented to the participants as stimuli. We are using MVPA, Logistic regression (multiclass, one-vs-one) and decoding targets are emotional expression with L2 regularization.


I have come across a problem regarding time decoding using a sliding estimator. As in my pre-processing pipeline, I have already split my epochs into test and training data sets and performed ICA and other pre-processing steps already.


My question regarding the sliding estimator is, I want to use separate test and training sets for fit and prediction and I am unable to do so using MNE as in a sliding Estimator one can only feed the full data, and then it split the dataset on its own. Can you please help me regarding this, how can I feed separate data sets and get the same analysis results (have the respective scores)? Is there any solution in MNE or should I go for scikit-learn?


I have also attached the chunk of the code. Please let me know if you have any further questions or something is unclear.


Best,

Ammara Nasim[cid:579a854d-31ae-4f62-a516-d19fdfab441e]
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20220728/9135b707/attachment-0001.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pastedImage.png
Type: image/png
Size: 29396 bytes
Desc: pastedImage.png
Url : http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20220728/9135b707/attachment-0001.png