# ICA Epoch Ordering **URL:** \ **Category:** Mailing List Archive (read-only)\ **Tags:** list-archive\ **Created:** [August 23, 2017, 5:01pm UTC](https://mne.discourse.group/t/ica-epoch-ordering/1284 "2017-08-23T17:01:29Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![system](https://global.discourse-cdn.com/free1/uploads/mne/original/1X/85cc6bd2b69cb698a166dc6d880fb550510d0144.jpeg) [@system](https://mne.discourse.group/u/system)\ **Post date:** [August 23, 2017, 5:01pm UTC](https://mne.discourse.group/t/ica-epoch-ordering/1284/1 "2017-08-23T17:01:29Z")
Hello,     Has a consensus been reached on whether it is better to run ICA on epochs or on raw data before epoching?     On one hand, the ICA will only be applied base on the information in the epochs, irrelevant artifacts will not contribute, but on the other if the epochs are just appended to each other there will be a discontinuity at the connection of each pair of epochs. I have been using a filter buffer period as well if there is smoothing done to correct for this when running ICA on epochs. Thanks, Alex Rockhill -------------- next part -------------- An HTML attachment was scrubbed... URL: [http://mail.nmr.mgh.harvard.edu/pipermail/mne\_analysis/attachments/20170823/87bbc9ef/attachment.html](http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20170823/87bbc9ef/attachment.html) ---
**Author:** ![system](https://global.discourse-cdn.com/free1/uploads/mne/original/1X/85cc6bd2b69cb698a166dc6d880fb550510d0144.jpeg) [@system](https://mne.discourse.group/u/system)\ **Post date:** [August 24, 2017, 3:25pm UTC](https://mne.discourse.group/t/ica-epoch-ordering/1284/2 "2017-08-24T15:25:19Z")
hi Alex, our ICA algorithm work sample by sample hence it does not matter if you have discontinuities or if you shuffle the time points. if ICA works on Epochs just use it. Alex