[Eeglablist] Inconsistent results clean_artifacts follow up

[Delorme, Arnaud]

Dear Darniele,

Would you mind submitting a GitHub issue so we can assess if there is still an issue with reproducibility?

https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_sccn_clean-5Frawdata_issues&d=DwIFAg&c=-35OiAkTchMrZOngvJPOeA&r=kB5f6DjXkuOQpM1bq5OFA9kKiQyNm1p6x6e36h3EglE&m=mYXzpgbn7EKV7D3HEpzC-CiQG2eyPSRII0WM5ZtujakztlCatt8gtKGKASU6Mfsw&s=R-OtAGJc1hv0jQ0IctfNZlPs1ZHpUXX9-UTxbijR0TY&e= 

Arno

> On Mar 4, 2022, at 9:47 PM, Daniele Scanzi via eeglablist <eeglablist at sccn.ucsd.edu> wrote:
> 
> Hi there,
> 
> I thought to follow up this thread here (
> https://sccn.ucsd.edu/pipermail/eeglablist/2020/015456.html) about
> inconsistent results using *clean_artifacts* to detect and remove bad
> channels.
> 
> I was processing a dataset and encountered the same problem. After
> investigating for a while, I observed that the results are inconsistent
> when:
> 
> 
>   1. The recording is short (less than 10 minutes recorded at 1000Hz
>   before downsampling at 250Hz). So, in general, when the data has less than
>   150.000 samples. The less the number of samples, the more inconsistent the
>   results are.
>   2. The cut-off for the low-pass filter is 0.1 (as used in ERP research).
> 
> 
> It seems that these two conditions need to be both met to run into
> inconsistencies. Furthermore, the two conditions appear to be related. In
> general, the lower the number of sample, the higher the cut-off at which
> inconsistencies appears.
> 
> As reported in the thread above, I suspect that the reason relies on the
> use of *rand()* in *clean_channels *(line 187). Setting rng('default')
> before calling *clean_artifacts* produces consistent results. Although they
> are not reliable as it is unsure whether the flagged channels are actually
> noisy or not.
> 
> I do not think there is much that needs to be done (maybe having a message
> in case someone is trying to use a short dataset?). I will try to open an
> issue on Github to introduce this. But I thought that having this thread
> here might help someone in the future looking on Google for this behaviour.
> 
> Thank you for your work,
> 
> Daniele Scanzi
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu