[Eeglablist] pop_eegfiltnew vs. eegfilt - transition bandwidth

[Mary MacLean]

Great! Thanks for your response. Lots of useful info, and I may have some follow up questions after I’ve followed some of your suggested steps.

I am wondering though, if eegfilt has such severe problems why is it even still available in eeglab? Or at least why isn’t there a more explicit warning not to use eegfilt (or at least not to use it with the default settings)?

This seems to imply that eegfilt is not necessarily inappropriate. 

Thanks,

Mary



> On Apr 16, 2017, at 8:43 AM, Andreas Widmann <widmann at uni-leipzig.de> wrote:
> 
> 
>> We’re replicating another lab’s experiment. They used eegfilt for their data, when we use eegfilt we are able to replicate their effects. However, when we use eegfiltnew we do not. 
>> We changed the transition bandwidth to .15 from .25 in eegfiltnew
> You changed that directly in the code? This is actually a bad idea. To only partly change the heuristic for default filter order may have adverse side effects. The .25 parameter only applies to cutoff > 8 Hz (and < Nyquist - 8 Hz) in eegfiltnew anyway. I suggest manually computing filter order from requested transition band width and directly using this filter order in the GUI or CLI. I posted the equation recently on the list.
> 
>> and the data looks to, again, replicate their effects. This would suggest that a narrower transition bandwidth is yielding better SNR in this case. 
>> 
>> I have two questions:
>> 
>> (1) I realize that the narrow transition bandwidth will increase ripple in the passband
> Where did you read that? This is incorrect. Transition band width is a function of filter order. Ripple is defined by the windowing function. See
> http://home.uni-leipzig.de/~biocog/eprints/widmann_a2015jneuroscimeth250_34.pdf <http://home.uni-leipzig.de/~biocog/eprints/widmann_a2015jneuroscimeth250_34.pdf>
> for a detailed introduction of the concepts.
> 
>> , but is there any other reason why .15 may be problematic vs. .25? 
> Not problematic, just requires higher filter orders.
> 
>> (2) Is there any other difference between eegfilt and eegfiltnew besides the transition bandwidth? I couldn’t really tell from my cursory comparison of the functions. 
> Almost everything. eegfiltnew uses a completely different backend. The frontend was maintained for backward compatibility.
> 
> There are two different versions of the old eegfilt function. The older firls-default versions until early 2012 were ok-ish with respect to transition band width but had severe other (related) problems. The later fir1-default versions reduced these problems but transition band width was now actually far from the requested. See
> http://journal.frontiersin.org/article/10.3389/fpsyg.2012.00233/full <http://journal.frontiersin.org/article/10.3389/fpsyg.2012.00233/full>
> for a more details.
> 
> I suggest directly comparing the frequency responses of old and new filter to see the effective (!) filter characteristics find out what makes the difference for your data.
> 
> Best,
> Andreas
> 
>> 
>> Thanks for your help!!
>> 
>> -Mary
>> 
>> _______________________________________________
>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html <http://sccn.ucsd.edu/eeglab/eeglabmail.html>
>> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu <mailto: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 <mailto:eeglablist-request at sccn.ucsd.edu>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20170419/0d35b0ab/attachment.html>