[FieldTrip] the right normalsation?

"Jörn M. Horschig" jm.horschig at donders.ru.nl
Thu Jul 4 10:39:00 CEST 2013


Dear Ye,

normalizing has different purposes. On the one hand, as Arjen pointed 
out, it is necessary to normalize in source space to get rid of the 
depth bias (well, alternatively you could normalize the leadfields). On 
the other hand, it makes averaging over subjects reasonable - otherwise 
differences in e.g. scalp conducitivity (for EEG) or headshape and size 
(for MEG) might lead to biasing an average in favour of some subjects. 
It is just the same reason as using a baseline when analysing ERPs/ERFs.
Furthermore, there is no correct way of normalizing. As I said, you 
could also normalize the leadfield rather than normalizing by 
conditions. I think the most important thing to remember is that
a) you need some contrast or
b) you need to normalize the leadfield

For a) it would be sufficient to use Post#1/Post#2, but I would rather 
not contrast two conditions normalized by two baselines. This gets 
really hard in interpretation (e.g. is the observed effect caused by a 
difference in baseline or a different in stimulus processing?). My 
favourite is log(Post#1/Post#2) or taking the z-score. Taking the 
logarithm has the advantage that extreme values get squashed nearer 
together, thereby reducing the influence of outliers. Z-scoring achieves 
a similar thing by explicitly normalizing by the variance.
I would suggest for you to create some synthetic signals or values and 
play around with different ways of normalizing to get a feeling for what 
you are doing and what influence this has. I test different cases (e.g. 
3: difference in prestim but no difference in poststim, no difference in 
prestim but difference in poststim and difference in prestim and 
poststim) and apply different normalizations/contrasts.

Good luck :)
Best,
Jörn


On 7/3/2013 9:35 PM, Frank Mei wrote:
> Hello FieldtripList,
>
> I am trying to differentiate brain areas responsible for two different 
> conditions using the method show in the 
> tutorial(http://fieldtrip.fcdonders.nl/tutorial/beamformer), and so 
> far I have tried to subtract condition# 1 minus the #2, and divided by 
> the average of the baseline period (pre-stimuli presentation), i.e., 
> (Post#1-Post#2)./(Average of Pre#1 and Pre#2). I think this division 
> is for normalisation purposes. Is this the right normalsation? What 
> normalization do you suggest to use? Is it necessary to normalise?
>
> thanks
> Ye
>
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip


-- 
Jörn M. Horschig
PhD Student
Donders Institute for Brain, Cognition and Behaviour
Centre for Cognitive Neuroimaging
Radboud University Nijmegen
Neuronal Oscillations Group
FieldTrip Development Team

P.O. Box 9101
NL-6500 HB Nijmegen
The Netherlands

Contact:
E-Mail: jm.horschig at donders.ru.nl
Tel:    +31-(0)24-36-68493
Web: http://www.ru.nl/donders

Visiting address:
Trigon, room 2.30
Kapittelweg 29
NL-6525 EN Nijmegen
The Netherlands

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20130704/0e3d9e9d/attachment-0002.html>


More information about the fieldtrip mailing list