[FieldTrip] ft_connectivityanalysis

[jan-mathijs schoffelen]

Hi Lieneke,

I fixed point 1. The code should be available tonight on the ftp-server and on google code.

Best, 

Jan-Mathijs

Begin forwarded message:

> From: jan-mathijs schoffelen <jan.schoffelen at donders.ru.nl>
> Date: July 19, 2011 8:50:21 AM GMT+02:00
> To: Email discussion list for the FieldTrip project <fieldtrip at donders.ru.nl>
> Subject: Re: [FieldTrip] ft_connectivityanalysis
> 
> Hi Lieneke,
> 
> Sorry that your scripts broke. I will look into it.
> 
>> 1. I get the following error: 'Error using ft_selectdata, selection of or averaging across channels in the presence of both label and labelcmb is not possible'. 
>> 
>> -> My output from ft_freqanalysis and input to ft_connectivityanalysis = 'powerandcsd'. Eric didn't have the problem when using 'fourier', but then you don't get the field 'labelcmb' in your output. 
>> -> I can't help getting both label and labelcmb from freqanalysis if choosing 'powerandcsd'. If I remove the field 'label' from the structure, the function gives an error missing the field 'label' elsewhere. So I need both fields 'label' and 'labelcmb'. Previously, this problem did not exist. 
>> 
>> Did something change here? This brings me to point 2. 
> 
> 
> Is there a particular reason why you want to stick with 'powandcsd'? Using 'fourier' in the output will give you full flexibility later on in your analysis. Admittedly, you don't get a 'labelcmb' in your output, but you can still specify cfg.channelcmb before calling ft_connectivityanalysis. My suspicion is that this will solve your problem. The reason for your bug is that we cleaned up the code a bit (to make it more transparent), and that we had to backward-support a large number of potential ways in which users are used to using the functionality in FieldTrip. Once again, I will try to solve this issue, but I strongly urge you (and the rest of the community) to start using 'fourier' as a method in ft_freqanalysis. Eventually it will make life much easier (for both user and programmer ;o) ). Yet, I can imagine that if you are under time pressure to finish your thesis you don't feel like changing your approach. Anyway, I have written some documentation on this issue here: 
> 
> http://fieldtrip.fcdonders.nl/faq/in_what_way_can_frequency_domain_data_be_represented_in_fieldtrip
> 
>> 2. I would like to run ft_connectivityanalysis for a selected group of channelpairs. Again, I get the aforementioned error. 
>> The same error I get if I run ft_freqdescriptives for a selection of channels (not all channels!, in contrary to my point 1... there I get the error both when I select channels as well as if I say channels = 'all'). Therefore I suspect that the first problem is caused by a bug in ft_connectivityanalysis, not in the more general function ft_selectdata that both ft_freqdescriptives and ft_connectivityanalysis make use of. Does that make sense?
> 
> Not really. What is the error you get when calling ft_freqdescriptives? As such ft_freqdescriptives shouldn't work on the bivariate metric (i.e. cross spectra), and should not worry about labelcmb. 
> 
>> 3. In addition, I would like to calculate the standard error of my coherence output (if I can get it again at some point :)) using the method 'jackknife' (as recommended by Eric). ft_connectivityanalysis still has the option to use jackknife in its code, it just doesn't do anything with the setting anymore. The same goes for ft_freqdescriptives (but having it in ft_connectivityanalysis would do for me!). 
> 
> Are you sure? As far as I know, jackknifing should still be supported by both functions. It is of course crucial that you retain your single trial estimates in the input data (i.e. calling ft_freqanalysis with cfg.keeptrials='yes').
> 
>> I hope the points I've made are a bit clear, otherwise please let me know. It's a serious problem for me not being able to run the coherence analysis, since I'm trying to finish my master thesis within a few weeks from now... So it would be great if somebody could help me! 
>> 
>> Thanks in advance :). 
>> 
>> Best, 
>> Lieneke
> 
>  Best,
> 
> Jan-Mathijs
> 
> Dr. J.M. (Jan-Mathijs) Schoffelen 
> Donders Institute for Brain, Cognition and Behaviour, 
> Centre for Cognitive Neuroimaging,
> Radboud University Nijmegen, The Netherlands
> J.Schoffelen at donders.ru.nl
> Telephone: 0031-24-3614793
> 

Dr. J.M. (Jan-Mathijs) Schoffelen 
Donders Institute for Brain, Cognition and Behaviour, 
Centre for Cognitive Neuroimaging,
Radboud University Nijmegen, The Netherlands
J.Schoffelen at donders.ru.nl
Telephone: +31-24-3614793

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20110720/3eac60ba/attachment-0002.html>