[Eeglablist] Baseline power when comparing ERSP in different conditions

[Stefan Debener]

Yano,

Re 2, the answer is yes. In my opinion, you should not treat data from 
different experimental conditions (you want to statistically compare) 
differently in your preprocessing/artifact-reduction pipeline. Make sure 
that data from all conditions go into one ICA (or be submitted to the 
same temporal filter, etc), and only after identification of 'bad ICs', 
data should be sorted by condition and could be compared. If you run two 
different pipelines for your different conditions you will never know to 
what extent your pipeline generated any observed condition effects.

Best,

Stefan


Am 29.01.20 um 11:30 schrieb Shumpei Yano:
> Dear Experts.
>
> This is Yano. I am a graduate school student in Japan and studying EEG
> during walking. I just started working with EEGLAB.
>
> I have a trouble and 3 questions about common power baseline when comparing
> ERSP in different conditions.
>
> 1. Can I compare ERSPs directly ?
>     I know it is better to use common baseline power. However, in my
> opinion, I can distinguish the degree of power increase (decrease) in a
> gait cycle even if I compare ERSPs directly (with each baseline power).
>
> 2. Did I conduct a bad pre-process ?
>       I pre-processed the data according to Snyder, 2015, in which movement
> artifact is removed by rejection of Independent Components (ICs) residual
> variance (RV) larger than 15% or dipoles outside the brain. Then, the
> number of ICs is different between conditions.
> Is this a cause of power difference between conditions and my problem?
> Should I align the number of ICs for conditions?
>
> 3. Is it necessary to normalize powers ?
>      Should I normalize the power of each condition? (For instance, dividing
> power spectrum by the total power of each condition)
>
> A detail is as follows.  I wonder if you could tell me the solution.
>
> I am comparing Event-Related Spectral Perturbation (ERSP) in different
> conditions.
> I divided EEG signals into epochs of -1 to 2 seconds from each right heel
> strike.
> When calculating ERSP, epochs were time-locked to single gait cycles (two
> consecutive right heel strikes) with "timewarp" option of "newtimef"
> function.
> Then, I have a trouble in common baseline power.
>
> I read EEGLAB list and learned it is necessary to calculate common baseline
> power (CBP) when calculating ERSP difference in conditions. In my
> understanding, in the case of comparing condition A with B, CBP is
> calculated as CBP = 10log A + 10log B (Q1). I calculated ERSP of one
> subject with a CBP. However, the output figures seems strange. Many
> frequency bands of a condition is red (strong ERS) and those of the other
> is blue (strong ERD).
> I also calculated the power spectrum for each condition and confirmed they
> have different total power (Q2, Q3).
>
>
> Thanks, Best Regards
> Yano

-- 
Prof. Dr. Stefan Debener
Department of Psychology
Neuropsychology	Lab
University of Oldenburg
D-26111 Oldenburg
Germany

Office: A7 0-038
Phone: +49-441-798-4271
Fax:   +49-441-798-5522
Email: stefan.debener at uni-oldenburg.de