[FieldTrip] using ft_megplanar with sinle-trial TFRs

[Lozano Soldevilla, D. (Diego)]

Hi Fred,

Thank you for let us know this documentation discrepancy now corrected (https://code.google.com/p/fieldtrip/source/detail?r=9813)

Regarding your question, if you're computing induced TFR (power estimates for each single-trial) the cfg.combinemethod='sum' (default) method in ft_combineplanar will sum the vertical and horizontal components of each channel. As power is positive (squared), 1) and 2) options will ended up in the same result (although 2) simplifies the channel representation).

Might be you already knew it but you can use fieldtrip functions to test these type of questions and figure out yourself (see tfr.png figure, second and third rows):

http://fieldtrip.fcdonders.nl/example/combineplanar_pipelineorder

Above I made a simulation that basically tested you question. In addition, I also simulate ERFs asking the very same question. I know you did not ask for it but as the ft_combineplanar documentation did not clarify this issue, I took the chance to test it. You'll see that with ERFs the order of the ft_combineplanar call in the pipeline matters (see erf.png figure and the wiki example url above). The reason is because ft_combineplanar ('sum' method) for raw and timelock time of data, squares the vertical and the horizontal planar components (nonlinear transformation). Then, for ERF computations, one should first compute single-trial average first and the compute the megplanar/combineplanar. 

This is very important because the alternative (combine planar at single trial and then average) is not an ERF: ft_combineplanar will square the signal and random noise (plus time jittered components) will be (positively) added to the grand-mean (wont be cancelled by the negative components of the signal, as ERF approach assumes).

I hope it helps.

best,

Diego



----- Original Message -----
> From: "Frédéric Roux" <f.roux at bcbl.eu>
> To: "FieldTrip discussion list" <fieldtrip at science.ru.nl>
> Sent: Monday, 15 September, 2014 12:09:24 PM
> Subject: [FieldTrip] using ft_megplanar with sinle-trial TFRs
> Dear all,
> 
> my question relates to the computation of combined planar gradients
> using the ft_combineplanar function.
> 
> Our data has been recorded with an Elekta Neuromag 306 channel system,
> and I would like to know if it is "safe" to compute combined gradients
> before calling ft_freqstatistics, if the input data contains single
> trial TFRs.
> 
> The reason why I am asking is because in the documentation of the
> function it says
> % Use as
> % [data] = ft_combineplanar(cfg, data)
> % where data contains an averaged planar gradient (either ERF or TFR).
> 
> However, the following tutorial suggests that this can be safely
> applied if
> the input data contains TFRs which have been computed with the
> cfg.keeptrials = 'yes' option:
> http://fieldtrip.fcdonders.nl/tutorial/cluster_permutation_freq
> 
> I am bit confused as both information (the fieldtrip documentation and
> the tutorial)
> seem to contradict each other.
> 
> The 2 options that are see are:
> 
> 1) compute TFRs with cfg.keeptrials = 'yes', then call
> ft_frestatistics, then call ft_combineplanar
> by using the output of ft_freqstatistics as input for the
> ft_combineplanar function.
> 2) compute TFRs with cfg.keeptrials = 'yes', then call
> ft_combineplanar, then call ft_freqstatistics
> 
> Could anyone with experience on this issue please let me know what the
> best way
> to proceed would be?
> 
> Fred
> ---------------------------------------------------------------------------
> 
> 
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

-- 
PhD Student
Neuronal Oscillations Group
Donders Institute for Brain, Cognition and Behaviour 
Centre for Cognitive Neuroimaging
Radboud University Nijmegen 
NL-6525 EN Nijmegen
The Netherlands
http://www.ru.nl/people/donders/lozano-soldevilla-d/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: tfr.png
Type: image/png
Size: 40910 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20140918/41ae82b0/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: erf.png
Type: image/png
Size: 65168 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20140918/41ae82b0/attachment-0003.png>