[FieldTrip] Wavelet advice

Sheraz Khan sherrykhan78 at gmail.com
Mon Aug 20 03:29:42 CEST 2012


You can not avoid the border effects they comes from the convolution.

Only possible solution which I sometimes use when the epochs are shorter
due to the design, is to do the wavelet  decomposition on
the continuous data, and then epoch it, this will avoid the borders from
the epoch.

Sheraz
Martinos Center
MGH/MIT/Harvard

On Sun, Aug 19, 2012 at 7:20 PM, Peter Goodin <pgoodin at swin.edu.au> wrote:

>  Thanks for all the replies so far.
>
>  Re-reading my previous question I appear to have been a bit vague to the
> overall problem.
>
>  The problem is I've used wavelets widths starting from 1 and increased
> to the default 7, but only a value of 1 gives me any usable data around
> 4Hz.
>
>  What confuses me is that Kaplan et al, 2011 (
> http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001267)
> were able to get theta band activity using epochs of 1s (200ms pre 800ms
> post) without apparent boundary effects at a width of 5, but when I tried
> the same thing (including zero padding), the results were not usable.
>
>  Again, any advice and explanations as to why 1. a value of 1 is not
> usable and 2. why I may be getting these boundary effects would be greatly
> appreciated.
>
>  Peter
>  ------------------------------
> *From:* fieldtrip-bounces at science.ru.nl [fieldtrip-bounces at science.ru.nl]
> on behalf of Sheraz Khan [sherrykhan78 at gmail.com]
> *Sent:* Monday, 20 August 2012 6:58 AM
> *To:* FieldTrip discussion list
> *Subject:* Re: [FieldTrip] Wavelet advice
>
>   Hi,
> You can always try variables number of cycles may be start with 3 and then
> go to 5.
> Sheraz
> Martinos Center
> MGH/MIT/Harvard
>  On Aug 19, 2012 7:17 AM, "Peter Goodin" <pgoodin at swin.edu.au> wrote:
>
>>  Hi Fieldtrip list,
>>
>>  I'm attempting to use wavelets to analyse some data collected with a
>> neuromag system (sample rate of 500Hz, pre-trigger period is 200ms, post is
>> 1.6 seconds) but am new to time frequency analysis. I'm interested in lower
>> frequency bins (~4 to 20 Hz). The config settings I've been playing with
>> are as follows:
>>
>>  cfg.method = 'wavelet';
>> cfg.channel = 'MEG';
>> cfg.keeptrials = 'yes';
>> cfg.foi = [4:1:20]
>> cfg.toi = [-.1:.01:1.5]
>> cfg.width = x
>>
>>  Using a default width of 7, I'm getting large boundary effects leading
>> to a period of ~300ms (400 - 700ms post trigger) of calculated data at 4Hz.
>> I've played with the width and have found that a width of 1 gives only a
>> small amount of boundary effect at the extreme edges and shows a large
>> increase in the frequencies I'm interested in at a time point which
>> corresponds quite nicely with the ERF data also analysed.
>>
>>  Having read the Tallon-Baudry (1999) article, material from the
>> tutorials and items from the mailing list, I understand that by using a
>> wavelet width of 1 I've biased my results towards higher temporal vs.
>> spectral resolution, but considering the low range of frequencies I'm
>> interest in, is this a problem? Does using a low width with my cfg.foi set
>> as it is just lead to a large amount of redundant data?
>>
>>  Advice on this subject would be greatly appreciated.
>>
>>  Peter.
>>
>>
>>
>> _______________________________________________
>> fieldtrip mailing list
>> fieldtrip at donders.ru.nl
>> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>>
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20120819/64adcd73/attachment-0002.html>


More information about the fieldtrip mailing list