[Eeglablist] Extract absolute power using ERSP

[Arnaud Delorme]

Dear Thang,

> 1. Is this the right way to obtain absolute power? Should I even look at ERSP?

Yes, you can look either at ERSP or the spectrum when computing power. The absolute scale might be different but they should be proportional (assuming you calculate them using on the same time range using the same parameters). In fact, for users without the signal processing toolbox, the spectrum is computed using by averaging ERSP power for all time points. The reason absolute scales are different are based on many considerations (power in dB/Hz computed with FFT or wavelet etc...).

You are right to set the baseline to NaN, this avoids computing the baseline and computes the absolute ERS (event related spectrum).
Do this transform first

ersp_abs = 10.^(ersp{1}/10);

to transform back to absolute power, then do whatever else you want to do with the data (I have no idea what you minus 1 and times 100 comes from in ClustERD_cond1 = (10.^(ClustErspData_cond1/10)-1)*100).

> 1a. This is related to the question above: I don't think I can compute the percentage change of Pfurtscheller's ERS/ERD as I have here in the last step above since there is no baseline. Is this correct? 

I do not have the detail about this method so I cannot really answer this question. When you have the raw power, you should be able to recompute whatever you need with it including percentage change if you need to.

> 2. Would I be better off looking at power spectra instead?


See 1. ERSP and spectrum should give similar results.

Best,

A. Delorme

On 7 Jan 2013, at 09:12, Thang Le wrote:

> Hi everyone,
> 
> I have a few questions regarding absolute power which, admittedly, is a topic that has been discussed at length on here. Nevertheless, I'm still not able to wrap my head around a number of things. I would great appreciate your comments or redirection to an earlier post in case I missed anything in my research.
> 
> In my STUDY design, there are three conditions. Due to the concerns of using different baselines when comparing these conditions, I have opted to analyze each condition with its own prestimulus baseline separately. In order to do this, my intention is to first extract the absolute power (i.e. set 'baseline', NaN). To give you an idea of what I'm doing, the following is how I precompute my data:
> 
> %ERSP with no baseline
> [STUDY, ALLEEG] = std_checkset(STUDY, ALLEEG);
> [STUDY ALLEEG] = std_precomp(STUDY, ALLEEG, {},'allcomps','on','recompute','on','interp','on','ersp','on', 'erspparams', {'cycles' [3 .5] 'baseline',NaN, 'freqscale' 'linear' 'timesout' 200 'freqs' [8 16] 'nfreqs' 18},'itc','on');
> 
> After this, I compute the stats for a cluster of electrodes of interest:
> 
> %Compute Statistics for Cluster, alpha frequencies only:
> STUDY = pop_erspparams(STUDY, 'condstats', 'on','freqrange',[8 12]);
> [STUDY erspdata ersptimes pgroup pcond pinter] = std_erspplot(STUDY,ALLEEG,'timerange', [-5100 1800], 'baseline', NaN, 'channels',{ 'PZ', 'OZ', 'POZ', 'P1', 'P2', 'O1', 'O2','PO3','PO4'}); I don't think the 'baseline' NaN is needed here since the data have been precomputed without the baseline.
> 
> To extract the ersp data (for the first condition), I use the following calculations 
> 
> %Take average of all frequencies:
> AllSsErspData_cond1 = mean(erspdata{1},1);
> %Take Average ERSP data for all electrodes in cluster
> ClustErspData_cond1 = mean(AllSsErspData_cond1,3);
> %Compute ERS/ERD (percent change)
> ClustERD_cond1 = (10.^(ClustErspData_cond1/10)-1)*100;
> 
> My questions are:
> 
> 1. Is this the right way to obtain absolute power? Should I even look at ERSP? As I understand ERSP measures the changes in amplitude of a frequency spectrum at a given point of time relative to a prestimulus baseline. Without the baseline being specified as in my case, I'm unsure how ERSP is calculated. There was a post (http://sccn.ucsd.edu/pipermail/eeglablist/2011/004229.html) which deals with this but I'm still a bit hazy on the explanation. 
> 
> 1a. This is related to the question above: I don't think I can compute the percentage change of Pfurtscheller's ERS/ERD as I have here in the last step above since there is no baseline. Is this correct? 
> 
> 
> 2. Would I be better off looking at power spectra instead?
> 
> 
> Thank you so much for your help.
> 
> Thang Le
> 
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20130112/10b680f7/attachment.html>